product lifecycle management using multi agent systems models

Available online at www.sciencedirect.com ScienceDirect Procedia Computer Science 103 (2017) 142 – 147 XIIth International Symposium «Intelligent Systems», INTELS’16, 5-7 October 2016, Moscow, Russia Product lifecycle management using multi-agent systems models V.O Karasev, V.A Sukhanov* Bauman Moscow State Technical University, 5, 2-ya Baumanskaya st., Moscow, 105005, Russia Abstract The article describes application cases of intelligent systems technologies in modern product lifecycle management system Ba sic terms and definitions of product lifecycle management (PLM) practices and intelligent systems enumerated Intelligent solutions based on PLM and logistic support analysis (LSA) methods proposed Multi-agent software system for PLM and LSA related tasks solving and automation presented © 2017Published The Authors Published Elsevier B.V © 2017 by Elsevier B.V by This is an open access article under the CC BY-NC-ND license Peer-review under responsibility of the scientific committee of the XIIth International Symposium «Intelligent Systems» (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the XIIth International Symposium “Intelligent Systems” Keywords: Intelligent systems; product lifecycle management; logistic support analysis; multi-agent software Lifecycle management Lifecycle management is one of the major challenges in modern complex technical products development and production Lifecycle management appliances allow significantly reduce costs and improve product Product lifecycle management – multilevel product characteristics control on all product lifecycle stages1 The main objective of lifecycle management is the effective implementation of programs and performance requirements specified in product development and minimizing the life cycle costs Controllable modifications of product, it’s production and exploitations systems are common life cycle management technics Methodical and informational support are necessary for effective lifecycle management Effective tools for solving lifecycle problems and challenges are Product lifecycle management (PLM) systems Fundamental PLM problem may be considered with control theory point of view It should be noticed, that PLM systems are distributed in time and space and cover all lifecycle stages One of most effective technologies for * Corresponding author Tel.: +7-906-066-9040 E-mail address: vkarasev@students.bmstu.ru 1877-0509 © 2017 Published by Elsevier B.V This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Peer-review under responsibility of the scientific committee of the XIIth International Symposium “Intelligent Systems” doi:10.1016/j.procs.2017.01.034 V.O Karasev and V.A Sukhanov / Procedia Computer Science 103 (2017) 142 – 147 intelligent distributed system development is multi-agent method Common structure of integrated PLM system presented at figure Multi-agent systems Multi-agent system (M.A.S) is computer system composed of multiple interacting intelligent agents within an environment Agents can interact with each other and passive environment Main concept of agent’s technologies effectively adopted in most frames of applicable and system programming, artificial intelligence scope and distributed systems One of fundamental formal description of M.A.S agents was introduced by L Gesser in his book “MACE: A flexible testbed for distributed AI research” In accordance to L Gesser agents described by following set of properties: x x x x x x Class – group of agents, identified by class name Name – every agent has unique name in class, so agent can be uniquely identified pair of string: class and name Role – description of role, executed by agent in class scope Skills – variety of agent abilities Goals – variety of goals, that agent tends to achieve Plans – roadmap of agent actions Fig Common PLM system structure Agents can communicate with environment by messages and each other throw unified information environment One of possible implementation of this environment is database with network access Primary feature of agent is ability to produce “actions” In this article presented multi-agent model of automated lifecycle management information system (ALMIS) In this model agents are different specialized automated workstations ALMIS is distributed multi-user interactive system, designed for effective lifecycle management of complex engineering products like armored vehicles Main goal of this system is to provide analytical reports and additional information for decision-maker Decision-maker 143 144 V.O Karasev and V.A Sukhanov / Procedia Computer Science 103 (2017) 142 – 147 detects problems in product lifecycle handling, figure out cost of whole lifecycle and estimate product exploitation infrastructure investments Most of existing PLM systems are very complicated despite of graphic interface and powerful embedded help system and require a large amount of effort for mastering from user M.A.S allow to build highly effective interactive system Agents not only receive and execute user commands, but exercise intelligent and active decisions Model development In the scope of ALMIS development the distributed M.A.S model was introduced Model characteristics could be researched with some of existing methodic and optimization technics Some of this methods and technics provided in3 In particular: x Peculiar logic constructions and operation for agent’s workflow description Name – every agent has unique name x Probabilistic approach Some system workflow stages claimed to be subject of stochastic effects Flagship approach to agents (and whole multi-agent system) workflow description is logic analysis Within this scope of analysis intentionalism logic, introduced by Cohen, P.R and Levesque, H.J4 should be pointed out This concept guides to describe agent behavior with in connection with its inner characteristics (like objectives, knowledge, etc.) However, logic analysis could not effectively resolve problem of dynamic system behavior verification M.A.S formal model was developed to investigate ALMIS dynamic behavior This model consists of standalone interactive subsystems and agents Developed model presented at figure Fig ALMIS M.A.S model V.O Karasev and V.A Sukhanov / Procedia Computer Science 103 (2017) 142 – 147 145 Presented at figure agents interact each other and environment throw shared consolidated information space ALMIS consists of some space-time distributed subsystems: x x x x x x LSA – Logistic support analysis EED – Electronic exploitation documentation EC – Electronic catalogues TEM – Technical exploration monitoring TEA – Technical-economic analysis SSCF – Service system costs forecasting These subsystems connected each-other by online links and integral data exchange system In presented model every agent is standalone ergatic system, consists of workstation and workstation operator One of this agent’s workflow logic presented at figure Classification and some functions of this agents provided further There are many ways to classify presented agents By parent system o by executed functions for example So agents may be qualified to some classification categories: Service agents x LSA-1 – LSA subsystem administration x … Data acquisition x x x x LSA-1 – LSA subsystem administration EED-2 – gathering documentation templates data TEA-2 – gathering product characteristics and exploitation system data SSCF-1 – gathering developer-exploitation organizations interaction data Data and knowledge generation x x x x SSCF-1 – gathering developer-exploitation organizations interaction data EED-3 – electronic exploitation documentation development EC-3 – electronic parts catalogues development SSCF-2 – service system infrastructure description generation Data and knowledge actualization x EED-4 – electronic exploitation documentation browsing and maintenance x EC-4 – electronic parts catalogues browsing and maintenance x TEM-3 –product exploitation database maintenance Analysis x LSA-3 – Product systems engineering analysis x TEA-4 – technical-economic analysis and report generation 146 V.O Karasev and V.A Sukhanov / Procedia Computer Science 103 (2017) 142 – 147 Fig Agent TEA-4 workflow Based on provided classification knowledge logistic flow presented at figure This methodology was introduced by Pashkin M.P at 20055 This methodology offers knowledge control techniques for adequate knowledge acquisition, integration and transport form distributed sources to the interested users (agents) at right time and context Users (agents) can make adequate decisions, based on this knowledge Fig ALMIS «knowledge logistic» flow For every agent in model was described its functions, objectives, workflows and links between agents and consolidated information space This data was used for system modelling For example, let’s provide some functions of agent LSA-3: x x x x x Peculiar logic constructions and operation for agent’s workflow description Name – every agent has unique name Check product data and logistic structure Create links between logistic and functional structures elements Produce FMECA (Failure mode effects and criticality analysis) Analyze product service tasks V.O Karasev and V.A Sukhanov / Procedia Computer Science 103 (2017) 142 – 147 x … Conclusion Provided list of functions is shortened, but allows to get an overview of agent’s functionality and complexity Every of agents provided in model can be decomposed to underneath M.A.S So provided model is recursive Development of recursive M.A.S system is very comprehensive task and it out of scope of this article But may be investigated in the future research Prototype of distributed information lifecycle management system was developed using the basis of provided M.A.S model of ALMIS This prototype was dynamically tested and afford to proof concept of application M.A.S models for lifecycle management system development Prototype confirmed the model adequateness and helped to formalize requirements for industrial commercial version of distributed multi-agent lifecycle management system References Sudov EV, Levin AI, Petrov AN, Petrov AV, Borozdin DN Analiz logisticheskoy podderzhki Teoriya i praktika [Analysis of logistic support Theory and practice] Moscow, Inform-Buro Publ., 2014, 260 p Gasser L, Braganza C, Hermann N MACE: A flexible testbed for distributed AI research Distributed Artificial Intelligence.— 1987 Zaytcev ID Multi-agent systems in modelling of social-economic environment: Research of behaviour and verifycation by Markov’s chains Dissertation Novosibirsk, 2014 Cohen PR, Levesque HJ Persistence, Intention, and Commitment MIT Press, 1990 Pashkin MP Research and development of multi-agent knowledge logistics system for decision support system Dissertation Saint-Petersburg 2005 147 ... system development is multi- agent method Common structure of integrated PLM system presented at figure Multi- agent systems Multi- agent system (M.A.S) is computer system composed of multiple interacting... Primary feature of agent is ability to produce “actions” In this article presented multi- agent model of automated lifecycle management information system (ALMIS) In this model agents are different... 147 detects problems in product lifecycle handling, figure out cost of whole lifecycle and estimate product exploitation infrastructure investments Most of existing PLM systems are very complicated