Characterization of polyacrylonitrile thermal stabilization process for carbon fiber production using intelligent algorithms

Abstract

Composite materials have widened their application range in recent years. The polymeric composite reinforced with carbon fibers can be described as a high-performance structural material which merges two important features: low weight and mechanical stability. Carbon fiber production which uses polyacrylonitrile as precursor is composed of many stages such as polymerization, spinning, thermal stabilization, carbonization, and surface treatment. Thermal stabilization is the critical stage of this production process, during which aromatic rings are generated, and therefore the main factor for the carbon fiber structure definition and thus for this material quality. A thermal stabilization model using intelligent algorithms was developed aiming a possible optimization of the production process and consequent cost reduction. This work was based on real experimental data obtained from a composite material production pilot plant. A qualitative analysis was initially performed using Self-Organizing Maps trained with variables of fiber production reagents and process. Thereafter, a supervised training with feedforward backpropagation neural network was used for a quantitative analysis. Based on this quantitative analysis, the carbon fiber thermal stabilization process was simulated, obtaining 2.98% and 2.48% mean errors relative to experimental results of Volumetric Density and FTIR Conversion index, respectively.