Abstract: Performance of a controller usually involves many attributes, including stochastic and deterministic ones. This paper proposes an assessment approach on controller performance. According to different demands of a specific control loop, an appropriate weight is assigned to every attribute index. Besides, considering the impact of different disturbances on assessment benchmarks, a BP network based process disturbance identification is introduced. Moreover, the assessment benchmark is established based on historical data and a multi attribute decision making method is employed to achieve a comprehensive attribute assessment result. Both numerical simulation and TE process experiment demonstrate the effectiveness of the proposed method.