TY - GEN
T1 - Causal design knowledge acquisition by constructing BBN through FCM
AU - Kim, Yun Seon
AU - Kim, Kyoung Yun Joseph
AU - Cheah, Wooi Ping
AU - Yang, Hyung Jeong
PY - 2008
Y1 - 2008
N2 - Managing design knowledge is an important concern for industry, including engineering. Engineering firms are facing pressures to increase the quality of their products, to have even shorter lead times and reduced costs. There is also a trend towards globalization resulting in complex supply chains and the need to manage teams that are not necessarily co-located. Design knowledge needs to be exchanged and accessed efficiently. Other motivations for managing design knowledge are to provide a trail for product liability legislation and to retain design knowledge and experience as engineering designers retire. Fuzzy Cognitive Map (FCM) is one of the main formalisms for modeling, representing and reasoning about causal knowledge. Despite the fact that FCM has been used extensively in causal knowledge engineering, there is a lack of methodology for the systematic construction of FCM. Although some techniques were used in the individual construction processes, these techniques were either not systematically documented or too specific to the problem at hand. FCM and Bayesian Belief Network (BBN) are two major frameworks for modeling, representing and reasoning about causal design knowledge. Despite their extensive use in causal design knowledge engineering, there is no reported work which compares their respective roles. This paper deals with three topics, which are systematic constructing FCM, a methodology for FCM-BBN conversion, and comparison FCM and BBN. BBN has a sound mathematical foundation and reasoning capabilities, also it has an efficient evidence propagation mechanism and a proven track record in industryscale applications. However, BBN is less friendly and flexible, and often very time-consuming to generate appropriate conditional probabilities. Thus, Fuzzy Cognitive Map (FCM) is used for the indirect knowledge acquisition, and the causal knowledge in FCM is systematically converted to BBN. Finally, we compare BBNs directly generated by domain experts and generated from FCM, with a realistic industrial example, a fuel nozzle for an aerospace engine.
AB - Managing design knowledge is an important concern for industry, including engineering. Engineering firms are facing pressures to increase the quality of their products, to have even shorter lead times and reduced costs. There is also a trend towards globalization resulting in complex supply chains and the need to manage teams that are not necessarily co-located. Design knowledge needs to be exchanged and accessed efficiently. Other motivations for managing design knowledge are to provide a trail for product liability legislation and to retain design knowledge and experience as engineering designers retire. Fuzzy Cognitive Map (FCM) is one of the main formalisms for modeling, representing and reasoning about causal knowledge. Despite the fact that FCM has been used extensively in causal knowledge engineering, there is a lack of methodology for the systematic construction of FCM. Although some techniques were used in the individual construction processes, these techniques were either not systematically documented or too specific to the problem at hand. FCM and Bayesian Belief Network (BBN) are two major frameworks for modeling, representing and reasoning about causal design knowledge. Despite their extensive use in causal design knowledge engineering, there is no reported work which compares their respective roles. This paper deals with three topics, which are systematic constructing FCM, a methodology for FCM-BBN conversion, and comparison FCM and BBN. BBN has a sound mathematical foundation and reasoning capabilities, also it has an efficient evidence propagation mechanism and a proven track record in industryscale applications. However, BBN is less friendly and flexible, and often very time-consuming to generate appropriate conditional probabilities. Thus, Fuzzy Cognitive Map (FCM) is used for the indirect knowledge acquisition, and the causal knowledge in FCM is systematically converted to BBN. Finally, we compare BBNs directly generated by domain experts and generated from FCM, with a realistic industrial example, a fuel nozzle for an aerospace engine.
UR - http://www.scopus.com/inward/record.url?scp=81155128032&partnerID=8YFLogxK
U2 - 10.1115/DETC2008-49900
DO - 10.1115/DETC2008-49900
M3 - Conference contribution
AN - SCOPUS:81155128032
SN - 9780791843253
T3 - Proceedings of the ASME Design Engineering Technical Conference
SP - 43
EP - 51
BT - 34th Design Automation Conference
PB - American Society of Mechanical Engineers (ASME)
T2 - ASME 2008 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference, IDETC/CIE2008
Y2 - 3 August 2008 through 6 August 2008
ER -