Fault detection in robotic swarm systems is imperative to guarantee their reliability, safety, and to maximize operating efficiency and avoid expensive maintenance. However, data from these systems are generally contaminated with noise, which masks important features in the data and degrades the fault detection capability. This paper introduces an effective fault detection approach against noise and uncertainties in data, which integrates the multiresolution representation of data using wavelets with the sensitivity to small changes of an exponentially weighted moving average scheme. Specifically, to monitor swarm robotics systems performing a virtual viscoelastic control model for circle formation task, the proposed scheme has been applied to the uncorrelated residuals form principal component analysis model. A simulated data from ARGoS simulator is used to evaluate the effectiveness of the proposed method. Also, we compared the performance of the proposed approach to that of the conventional principal component-based approach and found improved sensitivity to faults and robustness to noises. For all the fault types tested–abrupt faults, random walks, and complete stop faults–our approach resulted in a significant enhancement in fault detection compared with the conventional approach.