Working Condition Recognition Based on Transfer Learning and Attention Mechanism for a Rotary Kiln

Abstract

It is difficult to identify the working conditions of the rotary kilns due to the harsh environment in the kilns. The flame images of the firing zone in the kilns contain a lot of working condition information, but the flame image data sample size is too small to be used to fully extract the key features. In order to solve this problem, a method combining transfer learning and attention mechanism is proposed to extract key features of flame images, in which the deep residual network is used as the backbone network, the coordinate attention module is introduced to capture the position information and channel information on the branch of feature graphs, and the features of flame images obtained are further screened to improve the extraction ability. At the same time, migration learning is performed by the pre-trained ImageNet data set, and feature migration and parameter sharing are realized to cope with the training difficulty of a small sample data size. Moreover, an activation function Mish is introduced to reduce the loss of effective information. The experimental results show that, compared with traditional methods, the working condition recognition accuracy of rotary kilns is improved by about 5% with the proposed method.

An Approach Based on Video Image Intelligent Recognition for Water Turbidity Monitoring in River

“…Numerous studies have been conducted to address the issue of small sample learning. − Transfer learning has been widely promoted and applied as a solution, with pretrained models based on ImageNet being utilized to adapt the fully connected structure for classification problems. These studies have shown that pretrained model parameters exhibit high accuracy in the target domain . Similar studies have confirmed this phenomenon, indicating that pretraining on large data sets enables the model to summarize the key features of image data through multiple layers of neural network mapping, resulting in strong generalization performance.…”

“…These studies have shown that pretrained model parameters exhibit high accuracy in the target domain. 23 Similar studies have confirmed this phenomenon, 24 indicating that pretraining on large data sets enables the model to summarize the key features of image data through multiple layers of neural network mapping, resulting in strong generalization performance. This insight is particularly valuable for small sample data sets.…”