Due to the complexity of the combustion processes of airbag Inflators, neural networks are adopted for the inverse problem of specifying combustion parameters in the design of airbag inflators. Successful use of neural networks has been demonstrated in this study. During the training and verification phases of the use of neural networks, the predicted target vectors (i.e., the diameter and number of the gas-generation propellant pellets) are within 1.3% and 1.9%, respectively, of the training and testing vectors. The predicted pressure-time histories in the discharge task are also in excellent agreement with the training and testing pressure-time histories. Neural networks are also found to be able to predict well outside the training domain in this study. Based on the results from a parametric study, the values of neural network parameters for obtaining the best training quality are also summarized in this paper.