In psychological network models, constructs are described as systems of interacting components, each node representing a distinct cognition, emotion, or behaviour. In contrast to the latent variable model, where redundancy is favoured, network models require unique items since redundancy can potentially distort the network estimation. The following study presents a correlation method which detects redundancy in networks by comparing edge profiles when potentially redundant items are included separately. A simulation study shows that two redundant items, when included separately in the network, have very similar edge patterns with non-redundant nodes, while the core network structure remains stable. The proposed approach is tested under various conditions and succeeds in most cases. When measurement error is high and sample size is low, the performance of the proposed approach decreases. The method was also compared to the weighted topological overlap (wTO) method, and while both approaches were successful in identifying the redundant nodes, the proposed approach provided clearer separation between redundant and non-redundant items. These findings demonstrate edge-pattern correlations as a useful technique for enhancing the reliability and interpretability of psychological network investigations. They also suggest future possibilities for threshold improvement, multiple redundancy management, and extending robustness in noisy environments.