基于ISDAE模型的复杂工业过程运行状态评价方法及应用

工业过程的运行状态评价对保证产品质量及提升企业综合经济效益具有重要意义. 针对工业过程中存在强非线性、信息冗余以及不确定性因素影响而难以建立稳健可靠的运行状态评价模型问题, 提出一种基于综合经济指标驱动的稀疏降噪自编码器模型(Comprehensive economic index driven sparse denoising autoencoder, ISDAE)的复杂工业过程运行状态评价方法. 首先, 在SDAE (Sparse denoising autoencoder)模型中引入综合经济指标预测误差项, 迫使SDAE学习与综合经济指标相关的数据特征, 建立ISDAE特征提取模型. 其次, 将ISDAE模型所学特征作为输入训练运行状态识别模型, 级联特征提取模型和运行状态识别模型并通过微调网络结构参数获得运行状态评价模型. 另外, 针对非优状态, 提出一种基于自编码器贡献图算法的非优因素追溯方法, 通过计算变量的贡献率识别非优因素. 最后, 将所提方法应用于重介质选煤过程, 验证所提方法的有效性和实用性.

Operating Performance Assessment Method and Application for Complex Industrial Process Based on ISDAE Model

The operating performance assessment of industrial process is of great significance to ensure the product quality and improve the comprehensive economic benefits of the enterprise. In view of the problems of strong process non-linearity, information redundancy and the influence of uncertainty factors in the complex industrial processes that are not conducive to establishing a robust and reliable operating performance assessment model, a comprehensive economic index driven sparse denoising autoencoder model (ISDAE) based operating performance assessment method is proposed for complex industrial processes. Firstly, SDAE (Sparse denoising autoencoder) is forced to learn data features related to comprehensive economic indexes by introducing comprehensive economic indexes prediction error term and a feature extraction model based on ISDAE is established. Secondly, the features learned from the ISDAE model will be used as input to train the operating performance identification model, and then the feature extraction model and performance assessment model are cascaded and the operating performance assessment model is obtained by fine-tuning the neural network. Then, for the non-optimal operating performance, a non-optimal cause identification method based on the autoencoder contribution plot algorithm is proposed, and the non-optimal cause is identified by calculating the contribution rate of the variables. Finally, the proposed method is applied to the dense medium coal preparation process to verify its effectiveness and practicability.