Integration of experience based decision support in industrial processes

Abstract

ALSTOM Transport, Rue du Docteur Guinier, BP 4,65600 stmeac -France. email: holitiana.rakoto@transport.alstom.com, {hermosi, ruet)@eniti? A b m e T b t a raarch d m at presentlog the 6nt elcmenh of a study carried oat In eanjjoncdon with Alrtom Transport an the "Experience Feedback" and "Laron Learned" problenutic. The paper b divlded in thm wctlolu The 6nt d o n prmenh the Experience Feedbsck proceu and problcmtlc la Alatom. In order to rolvc this problemtic, we prota the second r d o n a dedicated modd and methodology for a ratlond Implementation and cbolcr of decLdon rnpport In lbc Erpcrknce Feedback p m u h F h d y , the lblrd d o n concern the explolt.Iion of Experience Feedback In operational p m u c z a dncriblng the r a y to mmeve p u t rxpcrience~ acenrdlng to Ita dmllvitg rrlm cnmnt r1hUtloa Keyw-Exprlcnce feedback, Dsirlon rupport, Indostrid procmea modelling, C u e B a d Rurodmg I"R0DUCnONIn order to face quick variations of their envimnment, the dynamics of evolution in industrial companies (organisation changes, people mobility and so on) has considernbiy increased. Therefore, the capitalisation of past experiences and the ability to inject the lesson learned into operational industrial processes has become a strategic issue. The "Experience Feedbnck" (KF) may be considered as a structured approach for cnpitalisntion and exploitation of knowledge obtained in past success and failures.This paper aims at prcscating the first elements of a study carried out in conjunction with Alstom Transport on the "Experience Fdhnck" and "Lesson h e d " problemtic.Alstom Transport designs industrialises and ensures the maintenance of s e v d high technology devices (such as power modules, command pla Iforms... ) of the traction p m of trains, metros or tramways. An essential k t o r for customer satisfaaion is the high level of reliability of products, which must be taken into ~cmunt since the design phase pasign For Reliability) and dnring the whole life cycle of the products.Until a recent period, the reliability quirement has bem ensured thanks to the d e q techuical expertise of specialists patrimony of the company. However, several fanors have modified the way to apprehend this problem. Indeed, work organisation, in time and space, o h have a vcry fast dynamics of evolution which makes more di5cult to apply shnred and durable technical des. For instance, the development of a knowledge management project introdures problem of information sharing benuem past and nmning whose intcllecNd asset clearly constitute3 a tshnical projects. The rn up of an Experience Feedback (EF) is essential in such case. This paper is divided in three sections. The fust section presents the EF problematic and mention several approaches that have bem suggested in the literature. The second section presents a dedicated model and methodology for a rational implementation of decision support in the EF process. The third section conccms the exploitation of the EF in oprrational proceues. especially the way to retrieve past experiences accor...

Fusion of possibilistic sources of evidences for pattern recognition

“…These degrees are the average of inter-attribute necessity and possibility degrees that take into account, if required, specific points of view, to adjust the similarity of each attribute between the patterns modeled by the tolerance function. Even if, sometimes necessity and possibility degrees are aggregated using the minimum or the maximum operator [16], this strategy does not work for our purpose. In those cases the value of the most dissimilar attribute will dominate, determining the final computation result, regardless of the other attributes, for which similarity values may be extremely close.…”

“…In the first case histograms are normalized before applying an appropriate probability-possibility distribution transformation on which our possibilistic similarity measuring can be carried out [16]. In the second case, an equivalent operation can be carried out even if the attributes are described by means of heterogeneous and imperfect information elements [3].…”

Case-based reasoning and system design: An integrated approach based on ontology and preference modeling

“…During the definition of a new experience, prior capitalized experiences are taken into account in order to be reused. Different approaches exist: experiential learning (Kolb, 1984), lessons-learned systems (Weber et al, 2001), EF loops (Faure & Bisson, 1999; Rakoto et al, 2002; Jabrouni et al, 2009, 2011), or trace-based reasoning (Cordier et al, 2009; Settouti et al, 2009). In order to support such a process, tools as CBR are suitable for aiding the definition of experiences, their capitalization, and their future reutilization.…”

BPRM Methodology: Linking Risk Management and Lesson Learnt System for Bidding Process

“…In order to carry out activities related to LLS (capitalization, processing and exploitation) [9], an experience vector (set of structured information eventually nested) has been proposed. It allows to record all relevant information to the implementation of a LLS towards BP risk management and is made of five elements: i) the "context" describing the current BP framework, ii) the "analysis" conducting feasibility studies and risk analysis, iii) the "solution" describing the technical and commercial offer, iv) the "deployment" presenting the product development as well as the encountered difficulties (if the offer is accepted), and v) the "closure" giving the final risk analysis and synthesis of the project.…”