Soft Sensors Using Heterogeneous Image Features for Moisture Detection of Sintering Mixture in the Sintering Process

摘要

In the sintering process, the moisture of the sintering mixture is important information for sinter quality control, and online accurate detection of mixture moisture is an issue that sintering plants have been eager to solve. In pursuit of such a goal, this article proposes a soft sensor using heterogeneous image features. First, a moisture detection system is designed and installed near the sintering mixture to capture the mixture image online. Second, inspired by the water watchers' experience of watching the mixture, morphological features, texture features, and color features related to mixture moisture are extracted from the mixture images. Then, a soft sensor model, which takes the extracted heterogeneous features as input, is proposed based on a weighted ensemble semisupervised stacked autoencoder (WE-S-SAE). Finally, the mixture moisture can be predicted by using the WE-S-SAE-based soft sensor. Experimental results demonstrate that the performance of the proposed soft sensor is better than that of the near-infrared moisture detector (NIMD) used in the sintering site, the root mean square error (RMSE) is reduced from 0.31 to 0.174, and the hit rate (HR) is enhanced from 46.2% to 77.2%.

关键词