Abstract: As satellite payload functions and structures become increasingly complex, higher requirements are being placed on equipment stability and reliability. To address the problem of the huge health characterization parameters of the payload equipment, which leads to the difficulty of health diagnosis, a nonlinear Principal Component Analysis (PCA) method is proposed. The method firstly constructs a topdown merged set of health characterization parameters for the payload equipment. In the end, on the basis of traditional PCA, nonlinear data fusion technology is used to obtain the key health characterization parameters, and the eigenvalues, eigenvectors, and ratios of eigenvalues in the resulting covariance matrices are studied in depth, and the results show that nonlinear PCA algorithms can bring better dimensionality reduction and eigenvalue cumulative ratio, which not only can maintain data quality, but also can reduce the number of eigenvalues. The results show that the nonlinear PCA algorithm can bring better dimensionality reduction and eigenvalue ratios, not only maintain the information of the differences between the data, but also preserve the information of the samples themselves, which effectively reduces the information loss caused by the data processing as well as the demand for the onboard load computation and storage capacity, indicating that the improved method shows significant advantages in the extraction of the health characterization parameters of the payload equipment.