The aim of this paper is to show the interest of the covariance analysis applied to measurement error in the particular case of the identification of a drying characteristic curve from experimental drying data. The modelisation of drying by use of the Drying Characteristic Curve (DCC) method is first presented with usual specifications (power function, critical moisture content …). The experimental procedure used to obtain drying curves and the data processing are detailled and analysed. Measurements errors are identified at the first step of the procedure and their effects on the estimation error of the exponent α of the power function are estimated. Three different methods for estimating α are presented under their matrix form: the least square method and two methods based on the «Gauss–Markov» or «Maximum likelihood» theorem, firstly under a simplified form suited if the estimation errors are uncorrelated and secondly under a complete form suited even if the estimation errors are correlated. These three methods are applied to experimental results obtained with ginger roots drying. The value of the exponent α of the power function and then the distances between the three corresponding theoretical drying curves (representing product water content vs. time) and the experimental points are studied. It is shown that in this particular application, the complete Gauss–Markov method leads to the better fitting and that the simplified Gauss–Markov method, since it is a priori non appliable in this case where errors are correlated, gives quite better results than the oridnary least squares method. The covariance matrices of the estimation errors of reduced water content, reduced drying rate and exponent α are also presented in order to show the correlations existing between the measurement errors of each variable during a drying cycle.
No takes yet. Share an insight, caveat, or question.
Jannot et al. (2002) studied this question.