In this paper, we propose an unsupervised ensemble clustering approach for fault diagnosis in industrial plants. The basic idea is to combine multiple base clusterings of operational transients of industrial equipment, when the number of clusters in the final ensemble clustering (P*) is unknown. In practice, a Cluster-based Similarity Partitioning Algorithm (CSPA) is employed to quantify the co-association matrix that describes the similarity among the different base clusterings and, then, a Spectral Clustering technique embedding an unsupervised K-Means algorithm is used to find the optimum number of clusters of P* based on Silhouette validity index calculation. The identified clusters allow distinguishing different operational behaviors of the equipment. The proposed approach is verified with respect to an artificial case study representative of the signal trend behavior of an industrial equipment during shut-down operations. The obtained results have been compared with those achieved by a state-of-art approach, known as Cluster-based Similarity Partitioning and Serial Graph Partitioning and Fill-reducing Matrix Ordering Algorithms (CSPA-METIS): the results show that the novel approach is able to identify the final ensemble clustering with a lower misclassification rate than the CSPA-METIS approach.