A SENSOR FOR ONLINE MEASUREMENT OF THE VISCOSITY OF NON-NEWTONIAN FLUIDS USING A NEURAL-NETWORK APPROACH

Non-Newtonian fluids are characterized by a nonlinear relationship between viscosity and shear rate. Typical examples are some emulsions. The graph of viscosity as a function of the shear rate is known as the rheogram or the theological chart. In this work, a very simple mechanical device with minimal moving parts has been used as a sensor for measuring both the viscosity and the shear rate. A mathematical model to measure the viscosity in Newtonian fluids has been confirmed. However, a mathematical model of the sensor for the case of non-Newtonian fluids is difficult because several variables defining the transducing properties are not independent and the measurement is made during transient dynamic conditions. In solving this problem, a neural network approach has been used. The input consists of two voltages representing the sensor response and the temperature. The outputs are the viscosity and the shear rate. The measurement system is made out of a sensor head, an electronic circuit for powering the sensor and signal conditioning, and a neural network software, based in the back propagation learning algorithm. The neural net has been trained using experimental data from laboratory viscometer and a simplified mathematical correlation relating the viscosity and input voltages. The sensor has been tested by measurement of the viscosity and the shear rate for emulsions of heavy crude oil (bitumen) and water. Good correlation between experimental data and the system output was observed after the neural net training. On-line tests of the sensor are being conducted.