染色机助剂智能配送系统的构建及实践

针对传统的染色机助剂配送系统配送误差大的问题，提出了一种基于推荐预停值和预计用时预测的多层全连接神经网络模型。首先，使用配送过程记录的数据训练网络模型；然后，将需要配送的数据代入训练好的网络模型进行计算，得到推荐预停值和预计用时，推荐预停值与经验预停值按照可变比例算法计算最终预停值，系统根据最终预停值决定配送阀的关闭时机，利用预计用时评估配送过程是否超时。最后，使用4种预停模式各进行1 000次的助剂配送验证实验。结果表明，采用神经网络模型预测模式的配送误差的标准差为23.8 g,平均绝对误差为16.1 g,优于其他3种预停模式的配送误差，取得了较好的助剂配送精度。

Construction and experiment of intelligent chemicals distribution system for dyeing machine

Aiming at the large distribution errors in traditional chemicals distribution systems for dyeing machines, a multi-layer fully connected neural network model was proposed based on the prediction of recommended pre-stop value and predicted time. The network model was firstly trained using data recorded in the distribution process, and the data to be distributed was fed into the trained network model for calculation so as to obtain the recommended pre-stop value and predicted time. The recommended pre-stop value and empirical pre-stop value were used to obtain the final pre-stop value according to the variable ratio algorithm. The system worked to determine the closing time of the distribution valve according to the final pre-stop value. The predicted time was used to evaluate whether the distribution process was timed out. Four pre-stop modes were used for chemicals distribution experiments over 1 000 times, and the results show that the standard deviation of distribution error predicted by network model is 23.8 g, the mean absolute error is 16.1 g. It is superior to the other three pre-stop modes with better chemical distribution accuracy.