Ash Accumulation State Identification for Infrared Compensation Images of Air Preheater Rotor Based on Deep Learning Method

Abstract

Ash plugging of the rotary air preheater widely used in large-scale power station often occurs and even reduces the efficiency of the boiler in sever cases. Therefore, a deep learning-based method was proposed for analyzing the evolution of ash accumulation for the infrared compensation images of the air preheater rotor. The sample data of the infrared compensation images of air preheater rotor was preprocessed, and the denoised image was transformed into the gray-level curve image, and the Gaussian filtering method was used for the image enhancement. Then, the gray-level co-occurrence matrix (GLCM) was established, the correlation statistics were calculated, and five different types of texture feature parameters of angular second moment (ASM) energy, contrast, entropy, inverse difference moment (IDM) and correlation were extracted. Finally, a deep belief network (DBN) model was established, which was trained with those preprocessed infrared images. The testing results show that the proposed method can not only detect effectively and monitor the ash accumulation of the air preheater rotor, but also predict the occurrence of ash blockage in advance, so as to guide the operators to optimize the operation of the ash blowing system and ensure the normal operation of the air preheater.

Keywords

• power station boiler
• rotary air preheater
• ash accumulation
• infrared compensation imaging
• texture feature
• gray-level co-occurrence matrix (glcm)
• deep belief network (dbn)