The focus of this paper is on improving the monitoring and control of end-use quality variables in batch nylon 6,6 autoclaves. Relatively few on-line measurements and frequent disturbances often result in a significant amount of variability in the product quality variables for this system. To lessen the sensitivity to these disturbances, a fundamental model of the nylon autoclave process is developed from first principles and is subsequently used to develop and test practical reactor control configurations. Various FID implementations are evaluated according to their robustness to typical process disturbances. The model is also used to formulate strategies designed to detect process disturbances and monitor the development of the quality variables on-line. The process monitoring results are then used in conjunction with inferential quality control strategies to improve robustness to typical process disturbances. (C) 1998 Elsevier Science Ltd. All rights reserved.