Predictive control concept for large-scale production systems (LSPS) postulates a formation of corresponding finite-dimensional “attractor” for both the controlled company and its immediate environment (LSPS-ensemble attraction channel). Realization of such process requires, first of all, determination of controlled system disposition within this group-attractor. The estimation of groupstability changes of the LSPS-ensemble becomes very important factor for predictive control especially in crises situations. In the paper a modified Poincaré recurrence-plots map method as implemented for the adequate predictive control is discussed. The “passive” experiment method and ergodic hypothesis for estimation of the group-stability changes are also considered. The group-stability of the Russian ferrous LSPS in 2008-2009 crises is analyzed in detail. The modified Poincaré recurrence-plots map method along with the “passive” experiment data enables us to reveal and estimate the process of the LSPS self organization in crises situation.