Network reliability evaluation of manufacturing systems by using a deep learning approach

Cheng Fu Huang, Ding Hsiang Huang, Yi Kuei Lin*, Yi Fan Chen

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

1 Scopus citations


A manufacturing system with reworking actions is constructed as a stochastic-flow manufacturing networks (SFMN) because components (arcs and nodes) are with multi-state capacity. Network reliability is a useful indicator of the performance of an SFMN. It is defined as the probability that that a SFMN can satisfy a given demand. However, the network scale becomes complex in the environment of Industry 4.0 and big data context. The algorithm YKLIN (Lin and Chang in Computers & Industrial Engineering 63:1209–1219, 2012b) cannot calculate network reliability in time for those large cases. For responding network reliability immediately, this paper utilizes an architecture of a deep neural network (DNN) to propose a prediction model for network reliability evaluation. The proposed prediction model can estimate network reliability with a small error (root-mean-square error (RMSE) = 0.0022) in the numerical case. Furthermore, compared to the algorithm YKLIN, the computational time is significantly reduced for a large tile manufacturing system with 14 production lines. In detail, the algorithm YKLIN takes 56.78 s for evaluating network reliability of each data point, whereas the proposed model only takes 0.02 s. The proposed DNN model provides a feasible and efficient approach to achieve network reliability immediately for the real-world manufacturing system in the industry 4.0 environment.

Original languageEnglish
JournalAnnals of Operations Research
StateAccepted/In press - 2022


  • Deep learning
  • Deep neural network (DNN)
  • Industry 4.0
  • Network reliability
  • Stochastic-flow manufacturing networks (SFMNs)


Dive into the research topics of 'Network reliability evaluation of manufacturing systems by using a deep learning approach'. Together they form a unique fingerprint.

Cite this