In the life evaluation of polyimide (PI) film modified by nano TiO₂, a small sample insulation life evaluation method based on gray neural network was proposed to solve the problems such as lack of failure data, life distribution cannot be determined uniquely, high cost, and long test time. Nano TiO₂ modified PI film was prepared by in-situ polymerization method, and the accelerated electrical ageing test was conducted on the films to obtain the failure data. The grey neural network was constructed and trained to obtain the extended data with the similar characteristics and changing rules with the original data. The least square method was used for parameter estimation, and the life evaluation results of different failure data samples, distribution models, empirical cumulative failure functions were analyzed and compared. The different life evaluation schemes before and after extending were analyzed by calculation example. The results show that the combination of Weibull distribution and mathematical expectation formula is more suitable for life evaluation of original sample size.The evaluation of insulation life using expanded failure data has a better effect, and the combination of inverse power function model, lognormal distribution, and median empirical formula is the most suitable for the life evaluation of the extended samples.