Monitoring the performance of conveyor system using radio frequency identification in manufacturing environment: a recurrent neural network and genetic algorithm-based approach

Abstract

A number of new approaches to address the identification issues have been proposed recently, but due to the highly integrated nature of passive radio frequency identification (RFID) tags, it is difficult to evaluate them in real-world scenarios. A recurrent neural network-based hybrid approach with training through genetic algorithm has been proposed to model the performance of the RFID system with received power at the reader in the radio propagation channel as the implementable performance index. Target system is a conveyor system delivering multiple products. A method to deploy RFID technology has been developed and illustrated for smoothening flow on a conveyor. Although various analytical models have been proposed earlier, they fail to accurately predict the performance of RFID system. Proposed method incorporates various factors presented in the industrial environment, while only a few are considered in the analytical model. Such an integrated approach is a genuine extension of a previous model where only neural network model was tested to embrace the system's performance. A comparative study has been carried out to establish the better performance of proposed approach. The model proposed may be helpful to aid in the research area of simulation of RFIDs on computer for reflecting numerous factors in modelling for RFID system performance without sacrificing predictability.

Keywords:

• RFID
• genetic algorithm
• conveyor systems
• supply chain management
• simulation
• neural network

Acknowledgements

The work described in this article was partially supported by the facilities provided by The Hong Kong Polytechnic University. The authors also thank the editor and the reviewers for their valuable comments and suggestions that have led to the substantial improvement of the article.