Refine
Labor/Institute
Keywords
- Artificial Intelligence (17)
- ITS (Intelligent Transportation Systems) (17)
- Fahrerassistenzsystem (15)
- Künstliche Intelligenz (8)
- Prozesssteuerung (6)
- Autonomes Fahrzeug (5)
- Bayesian Networks (4)
- OOBN (4)
- Hirntumor (3)
- Papierindustrie (3)
Year of publication
Document Type
- Conference Proceeding (19)
- Article (13)
- Part of a Book (3)
Reviewed
- ja (1)
We discuss a Root Cause Analysis (RCA) system implementing a probabilistic approach based on Bayesian inference for adaptive reasoning under uncertainties in industrial process operation. The proposed approach is model based and accumulates the process knowledge within the problem domain, which data is gathered and stored in XML-based information server. The Bayesian networks have been created automatically from the XML-structured data. The interconnection between XML-failure trees is handled as object oriented instances of Bayesian sub-networks within master-network covering the entire process and monitoring its overall condition, output quality and equipment effectiveness. We implement sequential on-line adaptivity of models' parameters to reflect changes in process operation. The system learning can be supervised by user feedback on the actual root cause. The general RCA methodology is applied to plate cutting in a hot rolling mill.
We discuss the use of a hybrid system utilizing Object Oriented Bayesian networks and influence diagrams for probabilistic reasoning under uncertainties in industrial process operations. The Bayesian networks are used for condition monitoring and root cause analysis of process operation. The recommended decision sequence of corrective actions and observations is obtained following the "myopic" approach. The BN inference on most probable root cause is used in an influence diagram for taking decisions on urgency of corrective actions vs. delivery deadline. The build-in chain of causality from root cause to process faults can provide the user with explanation facility and a simulation tool of the effect of intended actions.
In the presentation a total system is presented, making use of data reconciliation, different types of diagnostics with respect to sensors, control loops and processes. These are used as inputs to a root cause analysis system, optimization and advanced control, using among others MPC, model predictive control. The system is being implemented at Visy Pulp and Paper mill in Tumut, Australia.
We propose a methodology for Root Cause Analysis (RCA), allowing fast and flexible decision support for operators, maintenance staff and process engineers in pulp and paper industry. RCA can identify non-obvious process problems and is therefore a powerful complement to normal automatic control. The general methodology is applied to a continuous digester.
Kurzfassung Diese Arbeit stellt ein robustes wissensbasiertes Verfahren zur Lückenbewertung für Spur-wechselmanöver vor. Zur Modellierung wurden dynamische Bayes-Netzwerke eingesetzt und mit Hilfe von Lernalgorithmen die Erkennungsleistung verbessert. Die Testergebnisse zeigen eine sehr hohe Trefferquote.
Learning Style Classification by Using Bayesian Networks Based on the Index of Learning Style
(2023)
Conditional automated driving (CAD) systems (SAE level 3) will soon be introduced to the public market. This automation level is designed to take care of all aspects of the dynamic driving task in specific application areas and does not require the driver to continuously monitor the system performance. However, in contrast to higher levels of automation the "fallback-ready" user always has to be able to regain control if requested by the system. As CAD allows the driver to engage in non-driving-related tasks (NDRTs) past human factors research has looked at their effects on takeover time and quality especially in short-term takeover situations. In order to understand how takeover performance is impacted by different NDRTs, this paper summarizes and compares available results according to the NDRT's impact on the sensoric, motoric and cognitive transition. In addition, aspects of arousal and motivation are considered. Due to the heterogeneity of the empirical work and the available data practically relevant effects can only be attested for NDRTs that cause severe discrepancies between the current driver state and the requirements of the takeover task, such as sensoric and motoric unavailability. The paper concludes by discussing methodological issues and recommending the development of standardized methods for the future.
This paper describes a novel approach to situation analysis at intersections using object-oriented Bayesian networks. The Bayesian network infers the collision probability for all vehicles approaching the intersection, while taking into account traffic rules, the digital street map, and the sensors' uncertainties. The environment perception is fused from communicated data, vehicles local perception and self-localization. Thus, a cooperatively validated set of data is obtained to characterize all objects involved in a situation (resolving occlusions). The system is tested with data, acquired by vehicles with heterogenic equipment (without/with perception). In a first step the probabilistic mapping of a vehicle onto a fixed set of traffic lanes and forward motion predictions is introduced. Second, criticality measures are evaluated for these motion predictions to infer the collision probability. In our test vehicle this probability is then used to warn the driver of a possible hazardous situation. It serves as a likelihood alarm parameter for deciding the intensity of HMI acoustic signals to direct the driver's attention. First results in various simulated and live real-time scenarios show, that a collision can be predicted up to two seconds before a possible impact by applying the developed Bayesian network. The extension of this network to further situation features is the content of ongoing research.
In this paper we introduce a novel approach towards the recognition of typical driving maneuvers in structured highway scenarios and identify some of the key benefits of traffic scene modeling with object-oriented Bayesian networks (OOBNs). The approach exploits the advantages of an introduced lane-related coordinate system together with individual occupancy grids for all vehicles. This combination allows for an efficient classification of the existing vehiclelane and vehicle-vehicle relations in a traffic scene and thus substantially improves the understanding of complex traffic scenes. We systematically propagate probabilities and variances within our network which results in probabilistic sets of the modeled driving maneuvers. Using this generic approach, we are able to classify a total of 27 driving maneuvers including merging and object following.
