Optimizing twin-screw food extrusion processing through regression modeling and genetic algorithms

Response surface analysis has become a standard for characterization of extrusion experiments in recent years. While response surface experiments provide large amounts of useful data, the problem persists in how data can be used to successfully design specified products for a consumer. The use of genetic algorithms was explored as a potential tool that can help solve response surface data to identify extrusion conditions needed for desired product design. Response surface regression was conducted on five varieties of peas and the regression equations were used to create a way of measuring fitness in a genetic algorithm model routine. In doing so, extrusion conditions of screw speed and temperature for were successfully predicted for response factors (radial expansion, density, WAI, WSI, pressure, motor torque, SME, and color) of all the pea varieties with strong fitness (>0.90). Results suggest that optimization using genetic algorithms can have a beneficial impact selecting extrusion conditions.