Hybrid data model for the improvement of an ultrasonic-based gravity measurement system

This contribution describes the online data processing of raw data collected from a system that is to determine the density of beer during fermentation by measuring the ultrasonic velocity. The ultrasonic velocity in pure liquids is highly correlated with the media's density. Transferring this principle directly to a fermenting liquid is rarely possible, because it possesses huge interferences from other process quantities like the temperature or the carbon dioxide concentration. Using the beer fermentation as an example process, different online filter steps are described which eliminate the occurring influences. An artificial neural network (ANN), trained during the fermentation, compensates the temperature influence. A frequency analysis of the temperature eliminates existing disturbances caused by varying concentrations of carbon dioxide. An online estimation of the intrinsic parameters from an unstructured model using a simplex algorithm for parameter optimisation sustains the measured quantities and grants predicting abilities. The system was applied to an industrial plant. For the extract (In brewing practise density is mostly expressed as the concentration of a corresponding saccharose solution with the same density at 20 °C. This physical quantity is called extract.) determination in fermenting beer the relative standard deviation (RSD) compared to an offline standard reference method could be reduced from 0.49% to 0.14%. Thus, an enhanced process control strategy based upon the measured extract concentration is made possible.

[1]  Peter Hauptmann,et al.  Improved ultrasonic density sensor with reduced diffraction influence , 1998 .

[2]  K. Lang,et al.  Learning to tell two spirals apart , 1988 .

[3]  David Julian McClements,et al.  Ultrasonic pulse echo reflectometer , 1991 .

[4]  James R. Skorpik,et al.  On-line ultrasonic density sensor for process control of liquids and slurries , 1999 .

[5]  Fluid density and concentration measurement using noninvasive in situ ultrasonic resonance interferometry , 1992, IEEE 1992 Ultrasonics Symposium Proceedings.

[6]  F. Dunn,et al.  VHF/UHF range bioultrasonic spectroscopy system and method , 1995, IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control.

[7]  Tom Tollenaere,et al.  SuperSAB: Fast adaptive back propagation with good scaling properties , 1990, Neural Networks.

[8]  Laser Ultrasonic Instrumentation for Accurate Temperature Measurement of Silicon Wafers in Rapid Thermal Processing Systems , 1998 .

[9]  Martin A. Riedmiller,et al.  A direct adaptive method for faster backpropagation learning: the RPROP algorithm , 1993, IEEE International Conference on Neural Networks.

[10]  J. M. Hale.  Ultrasonic density measurement for process control , 1988 .

[11]  Antonio Delgado,et al.  Observation and control of the beer fermentation using cognitive methods , 1953 .

[12]  Jaroslav Votruba,et al.  Education of Beer Brewing Physiology and Technology Using PSI Simulation Language , 1995 .

[13]  William H. Press,et al.  Numerical recipes in C , 2002 .

[14]  Arthur E. Bryson,et al.  Dynamic Optimization , 1998 .

[15]  Motonao Tanaka,et al.  Phase correction on ultrasound speed measurement , 1998 .

[16]  Geoffrey E. Hinton,et al.  Proceedings of the 1988 Connectionist Models Summer School , 1989 .

[17]  U. Flitsanov,et al.  Nondestructive ultrasonic determination of avocado softening process , 1999 .