Due to the changes of external environment, unpredictable disturbances, etc., the industrial processes often have nonstationary characteristics. Assessment of operating performance for industrial processes with a hybrid of stationary and nonstationary variables is rather challenging since the statistical properties are time-variant. To solve this problem, a meticulous model for feature extracting to assess the operating performance of nonstationary processes is proposed in this paper. After distinguishing the stationary and nonstationary variables, the stationary variables are characterized by principal component analysis (PCA) to obtain the principal variation information, and the nonstationary variables are characterized by co-integration analysis (CA) to obtain the long term equilibrium relationship. Hence the inner features of both stationary variables and nonstationary variables are meticulously extracted. Next, to characterize the correlations between the stationary and nonstationary variables, a global model is established using the features obtained by PCA and CA. Since both types of features are stationary, the global model is built by PCA to extract the cross features of the stationary and nonstationary variables. Then an assessing strategy is developed based on the hierarchical hybrid model to online evaluate the process operation performance in detail. Finally, the application to a numerical example and a real copper flotation process illustrates the feasibility and efficiency of the proposed method.