Value-driven uncertainty-aware data processing for an RFID-enabled mixed-model assembly line

The use of radiofrequency identification (RFID) technology generates a high-volume, simple and unreliable data stream due to the technology's inherent unreliability. Such a data stream cannot be directly used for applications, as doing so would lead to inaccurate and unreliable result. In this paper, we propose a value-driven uncertainty-aware data-processing method that considers RFID detection reliability, timeliness and the throughput of an assembly line to characterize the potential benefits of RFID implementation in a mixed-model assembly system. The proposed method includes three components: a complex event processing system, a Bayesian inference model and a value-driven optimization model. We then demonstrate the use of the method by analyzing an automotive mixed-model assembly line. Some insights into management techniques are offered based on a comparison with several existing barcode implementations. The results of the method application also demonstrate that data uncertainty cannot be ignored in RFID cost-benefit analysis. Besides decreasing the cost of RFID technology, improving reading reliability and designing a sophisticated network can also offer significant benefits.