A methodology for the design of robust stable controllers for non-linear processes is proposed. The non-linear process is initially identified in the form of a non-linear ARMA polynomial. This model is then transformed into a non-linear state affine model which is especially convenient for robustness analysis and design. The model uncertainty is calculated from the non-linear terms of the non-linear state affine model. The uncertainty elements are shown to be a function of the input only and this facilitates the calculation of the uncertainty bounds. The technique is illustrated through a CSTR (continuously stirred tank reactor) control problem.