As an advanced visualization technique, electrical resistance tomography (ERT) has been used in dredging engineering in recent years. ERT can be used to measure and compute the solid-phase fraction and flow velocity in the pipe of any dredger in a non-intrusive way. Fault and anomality are inevitable in the long-term operation of ERT system, but the research on this phenomenon is insufficient. In this paper, we categorize all ERT faults to six classes. Various features from the ERT measurements are extracted. According to these features and relative class labels, a Choquet integral-based classifier is constructed. The use of Choquet integral rather than other techniques can effectively reflect the interaction and nonlinear relation among these features. Simulations and experiments were conducted in the dredging engineering experimental platform to verify the feasibility and effectiveness of the proposed method.