Complex industrial processes usually contain hybrid characteristics; that is, static and dynamic characteristics (SADC) exist between process variables, which bring great challenges to operating performance assessment. However, the existing achievements rarely consider the hybrid correlations. Although dynamic principal component analysis has explored this issue, it treats the static and dynamic variables indiscriminately, which artificially increases the dimension of modeling data to a certain extent. In view of the above problem, this paper proposes a distributed modeling scheme based on hybrid characteristics decomposition. First, an effective process decomposition algorithm that takes into account SADC is presented. Second, the suitable modeling strategy for each subblock is selected according to the strength of the SADC. Subsequently, the assessment results of different subblocks are merged through Bayesian inference, so that the final decision of triggering nonoptimal alarm is simplified. The proposed distributed principle component analysis-canonical variate analysis (D-PCA-CVA) algorithm more adequately mines the information contained in process data, thereby improving the ability of detecting nonoptimal status. Finally, the superior performance of D-PCA-CVA in operating performance assessment is verified through numerical example and gold hydrometallurgy process.