Research On The Detection Method Of The Crack On The Side Plate Of The Chain Grate Car

The steel industry is often used as an important indicator to measure the industrial level of a modern country.China's crude steel production accounted for 53.3%of the global total in 2019,ranking first in the world,mainly based on the grate-slewing machine with large output and stable product quality.The kiln pellet generation technology is mainly used.However,the structure of the grate trolley is complex,and the side plates will deform due to the alternating temperature changes,causing cracks or even partial breakage,which affects the sintering quality of the pellets.Therefore,it is necessary to check the damage of the side plates regularly.At present,the detection method based on manual inspection is timeconsuming and laborious.This paper uses image processing technology to design a crack detection method combining deep learning algorithm and computer vision technology,which provides a new idea for the field of crack detection in the side plate of the grate.The main research content of this paper is divided into two parts:side slab crack detection based on DeepLabv3+semantic segmentation model and calculation of the number and length of cracks based on image processing.The first is the production of the crack data set.This article uses the image video of the side plate of the grate machine taken on the spot in the industry,cuts into pictures according to the frame rate,and selects 1000 pictures with better definition and more obvious cracks for annotation,as the training data set.Then there is the choice of the crack detection algorithm model.The DeepLabv3+model developed by the Google team was chosen as the side slab crack detection algorithm.Use the well-labeled and resolved fracture data set for training,and use multiple prediction experiments to verify the optimal weight model.Finally,the number and length of the cracks are calculated.After the recognized image with the mask is processed by basic image processing methods such as HSV color space separation,Gaussian filtering,flooding and filling,the connected domain and skeleton based on morphology are extracted.Operate to obtain the skeleton information of the cracks,calculate the number and length of the cracks,and analyze the current damage of the side panels according to the set damage rules,give an alarm if necessary,and arrange maintenance work in time.This paper combines the crack detection of the side plate of the chain grate with the calculation of the crack parameters based on image processing technology,and realizes the identification of the cracks in the side plate of the trolley,to the calculation of the number and length of the cracks,and then based on this All aspects of judging the damage of the side panel.In the experiment,the accuracy of crack identification reached 99%,the accuracy of the calculation of the number of cracks was 94%,the accuracy of the calculation of the crack length was 96%,and the average error rate was 2.12%.The experimental results show that the method can effectively identify the cracks in the image,and it has certain reference value for modern steel companies to improve the quality of pellet sintering,extend the service life of machinery and equipment,guide industrial production,and improve corporate efficiency.