Selection of Signal Processing Techniques for Extracting Quantitative Indicators of Paste Quality from Extrusion Pressure Data

Signal processing techniques have been developed for extracting quantitative indicators of the performance of extruders processing soft-solid pastes from the data recorded by pressure transducers located near the extrusion die. Such techniques have been applied previously to isothermal ram and screw devices using retrospective methods. This article reports on the scope for application of retrospective algorithms, ideally suited to ram extrusion, to continuous systems. Novel Bayesian methods for the detection of outliers were found to be unsuitable for real-time analysis, prompting the development of a gradient-based algorithm. A number of fractal analysis techniques for quantifying the paste homogeneity were assessed and proved to be less robust than indicators based on coefficient of variance. Periodic variations related to acute, circumferential fracture were readily identified by a Bayesian model-based approach. Application to practical systems is illustrated by tests on several different paste materials following theoretical investigations on numerically generated data.