Integrated products–systems design environment using Bayesian networks. Issue 7 (3rd July 2017)

• Record Type: Journal Article
• Authors: Hanafy, Mohmmad; ElMaraghy, Hoda
• Is Part Of: International journal of computer integrated manufacturing. Volume 30:Issue 7(2017)
• Page Start: 708
• Page End: 723
• Publication Date: 2017-07-03
• Subjects: manufacturing systems -- Bayesian networks -- design environment -- structure learning -- k-most probable configuration algorithm; Computer integrated manufacturing systems -- Periodicals; 670.427
• DOI: 10.1080/0951192X.2015.1099072
• Languages: English
• ISSNs: 0951-192X

Abstract

Abstract : Providing new design environments to assist manufacturing engineers is essential to decrease products' lead time. Complex systems will be better designed and utilised to manufacture more products efficiently if systems–products' relationships are retrieved automatically and effectively. In this paper, a design environment using a Bayesian network is proposed. It inducts the relationships within the products' and machine's domains and maps the relationships between the two domains. The design environment incorporates Necessary Path Condition used in structure learning in a Bayesian network, estimation–maximisation (EM), k -most probable configurations algorithm, and d -separation concepts to understand and analyse these relationships, hence, it facilitates synthesising new systems and product. Two case studies are presented involving: (1) milling machines and their corresponding machined parts, and (2) tools' inserts used in grinding and their corresponding fixtures. In addition, a theorem is proposed, proved and discussed to justify the use of the Bayesian network for detecting the products–machines relationships. Results show that, unlike other design methodologies, the Bayesian networks can provide adaptable design environment by analysing the interactions between existing manufacturing entities such as machines/products and fixtures/inserts' specifications in a reverse engineering manner, without clearly identifying all the relationships between them a priori. The Bayesian network's inference capabilities are used to determine the most suitable machines/fixtures for new parts, and deduce the composite part/product that can be manufactured using newly acquired machines.