基于宽度回声状态网络的菜籽油加工参数自动决策方法研究

目的 实现菜籽油生产过程中加工参数的自动给定,研究基于人工神经网络的自动决策方法。方法 利用菜籽油加工过程的检测数据,建立一种宽度回声状态网络模型对加工参数与危害物的内在映射关系进行建模;在危害物含量要求下,利用此模型可实现加工过程参数的自动给定。结果 以脱臭工序为例的实验表明,所提方法能够有效利用已知变量自动计算出加工参数,宽度回声状态网络的计算精度优于其他几种典型循环神经网络模型。结论 所提方法可有效提升菜籽油加工过程危害物的自动控制水平,进而提升加工过程的科学性和规范性。

Study on automatic decision-making method of processing parameters of rapeseed oil based on broad echo state network

Rapeseed oil is an important part of bulk oil in China. Its processing technology and quality safety are very important. The parameters in the processing process of rapeseed oil have a direct impact on the reduction of a variety of hazards. Aiming at the decision-making problem of rapeseed oil processing parameters, an automatic decision-making calculation method based on machine learning is proposed in this paper. A broad echo state network model is established. The internal mapping relationship between processing parameters and hazards is modeled by using the measured data of rapeseed oil processing process. Then, under the guidance of the requirements of hazard content, the processing process parameters can be given automatically by using this model. Taking the deodorization process as an example, the experiment shows that the proposed method can effectively calculate the processing parameters automatically through the known variables, the calculation accuracy of the broad echo state network is better than several other typical cyclic neural network models, and can effectively improve the automatic control level of hazardous substances in the rapeseed oil processing process.