Towards fault detection of the operation of dairy processing industry tanks using Electrical Resistance Tomography

Abstract

Electrical Resistance Tomography (ERT) in monitoring milk tanks has previously been proven to have potential applications in the various stages of the milk powder production process for monitoring of crucial parameters. ERT provided 3-dimensional visualization of milk holding tanks for monitoring overall homogeneity or inhomogeneity, cream separation, aeration and object detection, which are parameters requiring control in the current milk processing industry. For the purpose of automatic detection and control of such faults, ERT-produced high dimensional data can be reduced to fewer dimensions covering most of the information.In this study, Principal Component Analysis (PCA) has been applied for the purpose of reducing the high-dimensionality of the ERT-produced data from such situations to lower dimensions holding most of the information. The reduced set of information is then used for the detection of whole and skim milk inhomogeneity, aeration and external object detection. A Matlab program has been developed which would use the above mentioned tools (ERT and PCA) to detect each of the mentioned faults in an opaque vessel which could be otherwise difficult to detect.