In view of the nonlinear industrial data disturbed by strong noise, a novel operating performance assessment method based on the stacked supervised denoising auto-encoders ( SSDAE) is proposed for complex industrial process. Firstly, a supervised denoising auto-encoder (SDAE) model is formulated, in which the performance grade labels are introduced to train the model. In this way, the SDAE can learn the characteristics closely related to the process operating performance and has strong ability to distinguish performance grades. Secondly, the SSDAE model is established by the stacking multiple SDAE model layer by layer and used to extract deep features which are closely related to the operating performance from the process data. Then, the deep features are used as the inputs of the SoftMax classifier. And the operating performance assessment model is achieved. Finally, the proposed method is applied to the hydrometallurgical process. Simulation results show that the assessment accuracy of SSDAE is up to 95% after the data are damaged by randomly setting zero in the proportion of 30% , which is obviously superior to other compared methods. Hence, the good performance and feasibility of the proposed method are verified under the condition of strong noise interference.