In this paper, an evolutionary neural networks model is proposed to predict the content of metal elements contained in marine diesel engine lubricating oil, by fusing genetic algorithms (GAs) and error back propagation neural network (BPNN) to offset the demerits of one paradigm by the merits of another. The input data of metal content was detected by spectrometric analysis. Genetic algorithms are used to globally optimize the weights and threshold of BP neural networks. Moreover, one case study was presented to illustrate the proposed method. The prediction accuracy of the novel method is compared with that of only BPNN method to illustrate the feasibility and effectiveness of the proposed method. The relative error on average of results is 1.52%, it can meet the precision request of state detecting in marine diesel engine.