Abstract: Quantitative evaluation is an important part of the simulation training of the Aviation Ammunition technical support. In order to realize the automatic evaluation of the simulation training, intelligent evaluation technology is introduced, and a prediction model based on Sigmoid regression is proposed. On the basis of analyzing the linear relationship of the sample data of the performance indicators, a subset of the characteristic indicators is selected as the input of the prediction mathematical model. In order to avoid the gradient descent method falling into the local solution problem, the gradient descent + PSO algorithm is presented. After testing the result samples, the algorithm can find the global optimal solution under the given precision. The prediction results have no over-fitting and under-fitting problems. In the actual practice, it is no longer necessary to input the subjective results of the examiners and experts, and the advantages of the computer-based simulation training automatic quantitative evaluation are brought into play.

Key words: aerial ammunition, simulation training; intelligent evaluating, particle swarm optimization, gradient descent method