Prediction of operating state for armored vehicle chassis engine and understanding the health state of engine in advance can effectively guarantee the combat capability of armored vehicle and extend its service life. A lightweight graph convolutional neural network (LGCN) was proposed to predict the operating state of an armored vehicle chassis engine. The quantitative analysis of Pearsons correlation coefficient was first carried out based on the characteristic data that affected the operating state of engine. The graph Laplacian matrix was constructed based on the quantitative structure of feature correlation. Then the Chebyshev polynomial was introduced to reduce the parameter amount and computation complexity of spectral domain graph convolution (GCN) calculation process. Finally, the operating state of armored vehicle chassis engine was predicted and analyzed based on the proposed lightweight graph convolutional neural network. The results show that LGCN can effectively realize the predictive analysis of operating state. The prediction results of multi-pattern recognition algorithm show that LGCN can obtain a classification accuracy of 98.75% and an F1-score of 98.31% with the best predictive stability.