Wu Y.-HHuang J.-YYao Y.-CTien Y.-JYu C.-JSHENG-DE WANG2022-04-252022-04-25202121682356https://www.scopus.com/inward/record.uri?eid=2-s2.0-85096389959&doi=10.1109%2fMDAT.2020.3036591&partnerID=40&md5=fbdbbe95c4d3b681c12798bdedeed0f0https://scholars.lib.ntu.edu.tw/handle/123456789/607280This article proposes a robust calculation method for the identification of anomalies in IC manufacturing test data while eliminating the need of large storage of raw measurements. Having the capability to process live streaming data is now the fundamental requisite for the successful realization of industrial Internet of Things (IIoT). The presented framework supports the in-stream analysis of data by the unsupervised incremental binning (UIB) technique as shown in one of the algorithms. UIB groups individual values into a small but sufficient maximum number of bins, such as default maximum 100 bins, with each bin recording observed sum and count. Upon a newly received value, UIB creates a new bin with sum value and count 1.anomaly detectioncollective anomalyequipment faultsrobust statisticssemiconductor test processesCost effectivenessCostsFault detectionCost effectiveLevel of useLive streamingOperational efficienciesReal timeSemiconductor testsIndustrial internet of things (IIoT)[SDGs]SDG9review10.1109/MDAT.2020.30365912-s2.0-85096389959