基于BP人工神经网络的蒸气云爆炸超压预测的研究

BP neural network study on predicting vapor cloud explosion overpressure

探讨用BP神经网络预测蒸气云爆炸超压的基本原理.重点分析影响蒸气云爆炸超压的各种因素,在此基础上利用BP神经网络算法,通过对80组不同反应活性气体在不同环境条件下爆炸的实验数据进行训练建立了预测蒸气云爆炸超压的网络模型,并对另外20组实验数据进行了爆炸超压预测.预测结果与实测结果基本吻合,说明该模型能够反映蒸气云爆炸超压与其多种影响因素的映射关系,可用于蒸气云爆炸超压预测.该方法为预测蒸气云爆炸超压提供了一条新途径,具有工程实际意义.

The basic principle of predicting vapor cloud explosion overpressure by BP neural network was discussed.The influencing factors of overpressure were analyzed emphatically.Based on this,a network model between overpressure and its influencing factors was established after a training process with 80 experimental data groups.Using this model,prediction was conducted for other 20 data groups.The results show the model is basically capable of formulating the relationship between overpressure and its influencing factors.It is an effective means for the predicting vapor cloud explosion overpressure.It is of certain practical significance and applicative value in engineering.