An investigation into control, optimization, and diagnostics of dilute-phase pneumatic conveying systems using artificial intelligence tools

Pneumatic conveying systems are widely used in chemical, mineral, metallurgical, and utility industries to transport bulk solids. Developing control and optimization strategies for such systems is challenging because of the variety of line configurations and equipment used. From discussions held with practitioners and consultants in this area, a need was identified for an intelligent system capable of characterizing controlling, and diagnosing pneumatic conveying systems. This work is a first step toward realizing such a system using artificial intelligence tools.; A set of dimensionless groups that have operational or physical significance in dilute phase pneumatic transport systems was identified using dimensional analysis. Data from conveying experiments carried out on two small to medium capacity test rigs using four different types of polymer pellets was transformed into such a dimensionless group format. Using a backpropagation neural network with faster training Levenberg-Marquardt method and early stopping technique to avoid overfitting, this data was used to build a predictive systems level model for dimensionless system pressure drop with an accuracy of around 4% (35Pa).; In a given pneumatic conveying system there exists a unique optimal condition where the amount of conveying gas is used to transport at a given solids flow rate, at the lowest possible system pressure drop and hence least energy. A fuzzy logic controller (FLC) was designed to identify and maintain such minimum pressure drop conditions in a test rig for conveying polyester pellets. The FLC was developed almost independently of the neural network model, using only qualitative information about the gradient of the pressure drop versus gas velocity curves and the current values of the gas velocity. The FLC was tested under static conditions (i.e.) its ability to identify and maintain consistent pressure minimum conditions. Under all tested conditions the FLC identified gas velocity corresponding to the optimal pressure minimum within ±0.2m/s in under 10 control steps.; A simple, reliable, inexpensive, and non-intrusive technique for metering solids flow was developed based on monitoring system wide pressure drops in dilute gas-solids flows. This meter was tested with a system conveying plastic pellets and was found to have an accuracy of ±15%.