探索用高维输入的神经网络对复杂工业生产过程的建模方法。针对网络输入变量维数较高的特点 ,提出一种 BP网络各权重独立训练的分散训练方法。该方法用附加大惯性项来协调各个权重的优化训练 ,运用非线性优化方法调节步长。与用普通的 BP训练方法相比 ,用该方法训练高维输入的 BP网络具有较快的收敛速度和较高的模型精度 ,较好地解决了实际生产过程的产品质量模型问题。 Exploring the Modeling Method of Complex Industrial Production Process Using High-dimensional Input Neural Networks. In view of the characteristics of high input variables in network, this paper proposes a decentralized training method of independent weight training in BP network. This method uses the additional large inertia to coordinate the optimization training of each weight and adjusts the step size by nonlinear optimization. Compared with the ordinary BP training method, the BP neural network trained by this method has faster convergence speed and higher accuracy of the model, and solves the problem of product quality model in the actual production process.