Diagnostic methodologies for nuclear power plants (NPPs) are usually based on mathematical models and generation of residuals. To avoid complicated, time-consuming, and costly diagnostic simulations of the physical phenomena in NPPs, an algorithm that determines a significant pattern for major transients is investigated. Coefficients of the transfer function between the observed parameters are used as the pattern features. The algorithm uses a recurring least-squares method known from the literature to determine the transfer functions. The case study includes 30 different scenarios in the primary and secondary systems. Each scenario produces its own significant recognized pattern. The RELAP5/MOD3.2 code is used to simulate the input data for the Krsko pressurized water reactor NPP. The algorithm recognizes the prepared scenarios, and it classifies them into groups.