Design And Development Of The Detection Software Of Semimolten Condition Of Fused Magnesium Furnace Based On Robust Stochastic Configuration Networks

Fused magnesium with high melting point,compact structure,strong oxidation resistance,chemical stability and other characteristics,which is important strategic raw materials.In China,the preparation of high grade fused magnesia is mainly recrystallized by electric arc furnace melting.Melting process can be divided into artificial feeding,melting and semi molten and other conditions.The feeding and smelting conditions are normal operating conditions,and the semi molten is an abnormal condition,it will affect product quality and production safety.At present,the plant to determine the condition of semi molten mainly by manual inspection,workers observe the shape of magnesium furnace flame,brightness,color and other characteristics,based on experience to determine whether there will be a semi molten condition.The manual inspection is easy to be influenced by workers experience and subjective factors,resulting in unstable product quality and low production efficiency of equipment.In the melting process of the fused magnesium furnace,there are significant differences in the color and shape of the flame in different working conditions,which can be used to judge the condition of the combustion by the video image automatically.Video image contains rich working condition information,through the flame form,the color and so on the characteristic analysis,realizes the semi molten condition perception,can help to reduce the worker to patrol the labor intensity,guarantees the safe production,has the remarkable application significance.In view of the problems in the identification of the electric magnesium furnace semi molten condition,this thesis adopts the newest research results of image feature extraction and stochastic configuration neural network and studies the condition identification of the electric melting magnesium furnace.The main work is as follows:1)The visual characteristics of the main working conditions of the electric melting magnesium furnace are analyzed by combining the process principle of the electric melting magnesium furnace and the experience of the operators.The sample data of various working conditions are selected from a large number of historical video data,and data preprocessing and working condition calibration are carried out to form a set of data set for modeling.2)Combined with the visual characteristics of semi molten,five quantized features are designed(Area of flame area,color average of whole image,flame region color average,flame brightness region color total and flame brightness);Video image feature extraction technology of electric melting magnesium furnace based on multivariable image analysis method.3)Using the characteristic data of the video segment of the labelling magnesium furnace as input,the classifier model of the semi molten condition is established by the robust stochastic configuration networks.The performance of the algorithm is compared with the BP neural network,RVFLNs and AdaBoost classifier based on decision tree.4)A demonstration software is designed and implemented based on the robust stochastic configuration networks to identify the semi molten condition of the electric magnesium furnace.The function of video preprocessing,the extraction function of the region of interest in the mouth,the feature extraction function of the flame image and the demonstration of the classifier recognition result are realized.