Diagnosis and Supervisory Control
- disco -
Research activities of the DISCO team
Objectives and Methodology
The aim of the DISCO team is to develop broad-spectrum methodological research in the field of diagnostics. The fundamental principle of a diagnostic process is to confront the uncertain/certain observation of a real system.
Machine learning for system diagnosis/prognosis
DISCO develops static classification methods with indicator extraction (normal, abnormal, healthy, sick, at fault, etc.) on populations of individuals using fuzzy logic and neural networks. A dynamic classification method, on the other hand, offers the possibility of learning dynamic behavioral states and detecting changes between these states, while also detecting new states. DISCO also contributes to the automatic learning of temporal models (such as chronicles) by exploiting temporal data mining techniques on system logs to discover temporally discriminable function classes.
Analysis of diagnostic models using formal methods
Formal study of diagnostic properties (diagnosability, identifiability, etc.) in dynamic systems (discrete, continuous, hybrid).
Development of diagnostic algorithms
The synthesis of diagnosers (diagnostic algorithms) is based on the existence of a model, a stream of observations and an objective level based on the knowledge derived from the model. This level of objective can range from the estimation of states to the synthesis of an explanation for the appearance of a fault.
Integrating diagnosis and prediction for predictive maintenance Diagnostic/prognostic integration for predictive maintenance
Prognosis is increasingly used in industrial applications, particularly for predictive maintenance. Diagnosis and prognosis are highly correlated: diagnosis determines the set of faulty components explaining the observed malfunctions, while prognosis determines the future state of the components.
Head
Scientific executive
Latest publications
2024
Journal articles
Conference papers
Other documents
Proceedings
@softwareversion
Preprints, Working Papers, ...
2023
Journal articles
Conference papers
Patents
Reports
Preprints, Working Papers, ...
2022
Journal articles
Book sections
Conference papers
Other documents
Reports
Preprints, Working Papers, ...
2021
Journal articles
Conference papers
@softwareversion
Reports
Preprints, Working Papers, ...
2020
Journal articles
Book sections
Conference papers
Other documents
Reports
DISCO team gathers permanent researchers from CNRS, INSA and University Paul Sabatier. DISCO is integrated into the model-based diagnosis (DX) and the fault detection and isolation (FDI) communities. DISCO team gathers researchers from both Automatic Control and Artificial Intelligence communities and addresses multidisciplinary research on system diagnosis based on the formalisms of both fields. By this multidisciplinary approach, DISCO team is unique in France.
Member of technical commitees:
- Hybrid Systems Technical commitee
- Model-based Diagnosis (DX) committee
- IFAC Safeprocess Technical committee
Association/Society member:
- IEEE Computer Society
- AFIA (Association Française en Intelligence Artificielle)
- PHM Society (Prognostics and Health Management)
Member of international and national editorial boards:
- International journal of intelligent manufacturing technologies
- Transactions of the Institute of Measurement and Control
- Artificial Intelligence
- Journal Européen des Systèmes Automatisés
Conference and Workshop organizations:
- 29th International Workshop on Principles of Diagnosis (DX-2018)
- Participation in the organisation of IFAC2017
- 26th International Workshop on Principles of Diagnosis (DX-2015)
- Workshop on Diagnosis Reasoning: Model Analysis and Performance (DREAMAP@ECAI-12)
- Workshop on Self-Star and Autonomous Systems (SAS@IJCAI-09)
Teaching, academic implications:
DISCO's professors and associate professors are involved in the creation of the following master courses
- INSA GEII: Distributed Systems and Big Data, Critical Embedded Computing Systems
- Univ. P.Sabatier: Electronics, Electrical Energy, Automatic: Robotics, Decision and Control and Real-Time System Engineering
DISCO's work is spread over several formations:
- (master) M2, ISTR, Univ. P. Sabatier, Toulouse
- (master) GEII, INSA, Toulouse
- (master) ENAC, Toulouse
- (PhD) EDSYS, LAAS, Toulouse
- (PhD) Summer Schools on Diagnosis, Spain
DISCO's research activities lead to the production of software and demonstrators.
Softwares
- COCD: unsupervised chronicle learning from a time sequence using data mining techniques. Alexandre Sahuguède, Euriell Le Corronc, Marie-Véronique Le Lann
- Diades: diagnostic tool for discrete event systems. Diades (DIAgnosis of Discrete Event Systems) provides a maintained set of algorithms and data structures for modelling, random generation, diagnosis and diagnostic verification in discrete event systems. Yannick Pencolé
- Dito: logical diagnostic tool based on CSP. This tool is a generic logical consistency-based diagnostic engine that transforms the diagnostic problem into a series of CSP problems (search for minimum cardinality diagnoses, search for minimum diagnoses). Yannick Pencolé
- Dyclee: dynamic classification tool. Dyclee (Dynamic Clustering algorithm for tracking Evolving Environments) implements a dynamic and self-adaptive unsupervised classification method with a novelty detection mechanism for the supervision of industrial systems. Renaud Pons and Louise Travé-Massuyès
- Locafleet: software for locating fleets of autonomous vehicles. Soheib Fergani
- HCDAM: Java-coded time pattern learning software that extracts the pattern common to a set of dated event sequences. Audine Subias, Euriell Le Corronc, Louise Travé-Massuyès
- Hydiag: diagnostic tool, active diagnosis and prognosis on hybrid systems. HyDiag (Hybrid Diagnoser) software is software developed under Matlab designed to simulate, diagnose and predict hybrid systems (model-based methods) by discrete abstraction of continuous dynamics and using Diades. Elodie Chanthery, Yannick Pencolé, Pauline Ribot, Louise Travé-Massuyès
- Hymu: diagnostic/prognostic integration tool. This software implements Hybrid Particle Petri Nets for monitoring the health state of a hybrid system (fault and degradation diagnosis and aging prognosis). Elodie Chanthery, Pauline Ribot.
- MaxPlusDiag: tool for detecting and locating time lags in time-discrete event systems based on algebraic theory (max,+). Exploitation of this technique as part of the monitoring of an automated production line in relation to its production objective. Claire Paya, Yannick Pencolé, Euriell Le Corronc
- Innograde: grade diagnosis assistance for breast cancer for anapathologists: Toulouse Tech Transfer depot, 2014. Marie-Véronique Le Lann
- P3S: fuzzy classification tool, Toulouse Tech Transfer repository. The P3S software (Process Sensor Selection and Situation assessment) uses fuzzy logic-based classification techniques for the selection of descriptors/sensors. Marie-Véronique Le Lann.
Demonstrators
- Prognospice: prototype demonstrating the diagnostic/prognostic integration of a pressure regulation failure by calculation methods with limited uncertainties. Prototype delivered to LIEBHERR (CORALIE project). Renaud Pons, Yannick Pencolé, Louise Travé-Massuyès, Pauline Ribot, Carine Jauberthie.
THESIS / HDR
2024
Louis Goupil, Thèse: Apprentissage machine guidé par des connaissances pour le diagnostic
2023
2022
2021
2020
Valentin Bouziat, Thèse: Gestion des aléas dans un système multi-robots
2018
2017
John William Vásquez Capacho, Thèse: Gestion d’alarmes basée sur des chroniques
2016
2015
DEPARTMENT
Jobs / Interships
REJOINDRE
Notre équipe de recherche
Pour plus d’informations sur les offres d’emploi, vous pouvez contacter