CHRONICLE BASED ALARM MANAGEMENT

John William Vásquez Capacho 1
1 LAAS-DISCO - Équipe DIagnostic, Supervision et COnduite
LAAS - Laboratoire d'analyse et d'architecture des systèmes [Toulouse]
Abstract : This thesis work was carried out in the framework of a co-tutelle between INSA, Toulouse, and the University of the Andes, Colombia, with financial support of Colciencias. This work is motivated by the need of the industry to detect abnormal situations in the plant startup and shutdown stages. Industrial plants involve integrated management of all the events that may cause accidents and translate into alarms. Process alarm management can be formulated as an event-based pattern recognition problem in which temporal patterns are used to characterize different typical situations, particularly at startup and shutdown stages. In this thesis, a new approach for alarm management based on a diagnosis process is proposed. Considering the alarms and the actions of the standard operating procedure as discrete events, the diagnosis step relies on situation recognition to provide the operators with relevant information about the failures inducing the alarm flow. The situation recognition is based on chronicles that characterize the situations of interest and are learned automatically. The chronicles are learned from representative event sequences obtained by simulation and given as input to an extended version of the Heuristic Chronicle Discovery Algorithm Modified (HCDAM). HCDAM has been extended in this thesis to account for expert knowledge in the form of specific temporal restrictions. A hybrid causal model of the process is used to verify the input event sequences and to explain and provide semantics to the learned chronicles. The Chronicle Based Alarm Management (CBAM) methodology proposed in this thesis involves different techniques to take the hybrid aspect and the standard operational procedures of the concerned processes into account. Compared to other approaches of alarm management, this approach uses information about the procedural actions related to the continuous variables behavior in a formal diagnosis process. Specific information is obtained in each step of the CBAM methodology, and it is summarized in three steps: 1. Step 1: Event type identification From the standard operating procedures and from the evolution of the continuous variables, this step determines the set of event types in startup and shutdown stages. 2. Step 2: Event sequence generation From the expertise and an event abstraction procedure this step determines the date of occurrence of each event type for constructing the representative event sequences. A representative event sequence is the set of event types with their dates of occurrence that can be associated to a specific scenario of the process. This step concludes verifying the representative event sequences using the hybrid causal graph. 3. Step 3: Chronicle database construction From the representative event sequences and temporal restrictions of each scenario, this step determines the chronicle database using the extended HCDAM algorithm. The proposed framework for alarm management is illustrated with two case studies representative of the petrochemical field.
Document type :
Theses
Automatic Control Engineering. INSA Toulouse, 2017. English
Liste complète des métadonnées

https://hal.laas.fr/tel-01660820
Contributor : Christine Fourcade <>
Submitted on : Monday, December 11, 2017 - 2:01:32 PM
Last modification on : Thursday, January 11, 2018 - 6:26:29 AM

File

VASQUEZ John William.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01660820, version 1

Citation

John William Vásquez Capacho. CHRONICLE BASED ALARM MANAGEMENT. Automatic Control Engineering. INSA Toulouse, 2017. English. 〈tel-01660820〉

Share

Metrics

Record views

9

Files downloads

5