When a disturbance occurs in a complex large-scale system, it may affect downstream equipment and several other process variables to evolve into a larger risk. The connectivity of process equipment may make it difficult to identify the propagation path of the disturbance. Understandably, the root cause identification of widespread disturbances gets its share of attention from researchers for remedial action but the prediction of the propagation path to prevent widespread disturbance is often overlooked. The scale-reasoning based risk propagation analysis method is proposed in this paper to predict the probable propagation path so that corrective actions can be taken in time to avoid further loss. By dividing the spatial scale of a complex production system, the approach uses transfer entropy to find the causal relationship between process variables and establish the risk propagation scale-reasoning model in the form of a causal map; then the risk propagation searching method, based on kernel extreme learning machine, is used to forecast the risk propagation path. An actual industrial case is analyzed to illustrate the effectiveness of the proposed method. (C) 2018 Institution of Chemical Engineers. Published by Elsevier B.V. All rights reserved.