Accurately detecting the change points of work mode is crucial for identifying the behavioral intentions of multi-functional radar (MFR). However, the intercepted MFR pulse sequence is filled with various noise, generating a large amount of measurement error, spurious pulses and lost pulses, making it difficult to locate change point positions. A window-sliding change point detection (Win-CPD) algorithm is applied in this paper, which detects change points by computing the discrepancy between two adjacent windows sliding along the MFR sequence, and discovering the peak in the discrepancy curve when the two windows cover different segments. Experiment results have verified the effectiveness and superiority of this algorithm.