Application of ultrasound and neural networks in the determination of filler dispersion during polymer extrusion processes

Mineral filler dispersion is important information for the production of mineral-charged polymers. In order to achieve timely control of product quality, a technique capable of providing real-time information on filler dispersion is highly desirable. In this work, ultrasound, temperature, and pressure sensors as well as an amperemeter of the extruder motor drive were used to monitor the extrusion of mineral-filled polymers under various experimental conditions in terms of filler type, filler concentration, feeding rate, screw rotation speed, and barrel temperature. Then, neural network relationships were established among the filler dispersion index and three categories of variables, namely, control variables of the extruder, extruder-dependent measured variables, and extruder-independent measured variables (based on ultrasonic measurement). Of the three categories of variables, the process control variables and extruder-independent ultrasonically measured variables performed best in inferring the dispersion index through a neural network model. While the neural network model based on control variables could help determine the optimal experimental conditions to achieve a dispersion index, the extruder-independent network model based on ultrasonic measurement is suitable for in-line measurement of the quality of dispersion. This study has demonstrated the feasibility of using ultrasound and neural networks for in-line monitoring of dispersion during extrusion processes of mineral-charged polymers. (c) 2005 Society of Plastics Engineers.