Recognition And Prediction System For Combustion State Of Rotary Kiln Based On Deep Learning

As the core equipment in the production process,rotary kilns are widely used in various industrial applications,such as incinerators,cement production,and steel production.The operating efficiency of the kiln depends on various parameters,such as inclination,temperature,speed,material flow rate and discharge rate.It is essential to keep the rotary kiln fully and stably burned to keep the temperature constant and improve product quality.Effectively identifying and predicting the combustion state in the rotary kiln is a very important and challenging problem in industrial production.However,the traditional image processing-based method requires a lot of preprocessing of the data in the early stage,and the accuracy is limited.This paper presents a system for identifying and predicting the combustion state of a rotary kiln based on deep learning.The proposed neural network architecture implements an end-to-end model of the output that directly draws upon input data,thereby eliminating the need for traditional complicated feature extraction procedures.The proposed convolutional neural network(CNN)can quickly and accurately identify the combustion state of the rotary kiln.In addition,the proposed convolutional recurrent neural network(CRNN)combines the advantages of convolutional neural network(CNN)and recurrent neural network(RNN)and can effectively predict the combustion state in the rotary kiln.The deep learning model inputs flame images into CNN and CRNN respectively to simultaneously recognize and predict the combustion state in the kiln.In order to improve the accuracy of CRNN network prediction,SC-III algorithm is introduced in the convolution layer.When the number of neural network nodes is small,the SC-III algorithm can achieve the best performance and effectively optimize the activation functions such as sigmoid and tanh.At the same time,the SC-III algorithm can randomly configure input weights and biases in the neural network,thereby enhancing the learning ability of the CRNN model.The main innovations are as follows:1.A deep learning method is proposed to identify the combustion state of the rotary kiln.The flame combustion video captured from the CCD camera is extracted into the convolutional neural network by frame image data to generate a three-dimensional prediction,corresponding to three combustion states(under-combustion,normal-combustion,over-combustion),the combustion state with the highest score is predicted It is the state of combustion in the rotary kiln at this time.2.A novel neural network architecture is proposed to predict the combustion state of the rotary kiln.The convolutional recurrent neural network first sends flame imagesequences with different window widths to the convolutional layer for convolutional feature extraction,and then enters each column of the obtained feature sequence diagram as a time series into the recurrent layer to obtain the prediction sequence,and finally passes The softmax classifier of the prediction layer predicts the combustion state.The collected data and simulation experiments show that in the proposed method,the convolutional neural network can quickly and accurately identify the combustion state of the rotary kiln.At the same time,CRNN can also effectively predict the combustion status in the kiln.Experimental results show that the method is effective and robust,and has great industrial application potential.