Most business processes in real life are not strictly ruled by the information systems that support them. This behavior is reflected in the traces stored by information systems. It is useful to diagnose in early stages of business process analysis. Process diagnostics is part of the process mining and it encompasses process performance analysis, anomaly detection, and inspection of interesting patterns.The techniques developed in this area have problems to detect sub-processes associated with the analyzed process and framing anomalies and significant patterns in the detected sub-processes. This proposal allows to segment the aligned traces and to form representative groups of sub-processes that compose the process analyzed. The tree of building blocks obtained reflects the hierarchical organization that is established between the sub-processes, considering main execution patterns. The proposal allows greater accuracy in the diagnosis. Based on the findings, implications for theory and practice are discussed.