The on-line diagnosis can be defined as a sequence of operations designed to detect, to isolate an anomaly and to mitigate its impact before it affects the inspected system. In the context of the project ALPES 2 (local accidents – protection and monitoring, in English) of the GIS 3SGS (group of scientific interest – surveillance, dependability, security of large systems, in English), the statistical methods for the early detection of a Total Instantaneous Blockage (TIB) occurred in a single subassembly of a sodium-cooled fast reactor have being studied. This work is carried out within the framework of model-based detection methods. The goal of this paper is to study the feasibility of a sequential approach to detect an abnormal local temperature rise of neighboring subassemblies due to a TIB while respecting the probability of missed detection, the probability of false alarm and the required time-to-alert. The adaptive statistical estimation/detection method is based on the measurements taken from ThermoCouples (TCs) located above the subassemblies. The aim is therefore to provide a complementary tool to improve the early detection of a TIB in the context of nuclear safety.