Dissertation
Dissertation > Industrial Technology > Metallurgy and Metal Craft > Metal pressure processing > Rolling > Rolling mill machinery equipment > Rolling automation

Research of Cold-rolled Steel Sheet Surface Quality On-line Monitoring System

Author QinZhenXing
Tutor ChenXueBo
School Liaoning University of Science and Technology
Course Control Theory and Control Engineering
Keywords Cold-rolled steel sheet Monitoring system RBF Neural network Defect recognition
CLC TG334.9
Type Master's thesis
Year 2012
Downloads 56
Quotes 0
Download Dissertation

With the increasing pace of cold-rolled steel sheet production speed, the traditional manual visual monitoring of surface quality of steel production has been difficult to keep up the current pace, and this method has many shortcomings,such as low inspection percentage,bad real-time performance,low detected degree ofconfidence,ete. so steel companies begin to increase emphasis on researching and developing cold-rolled steel sheet surface quality on-line monitoring system. As neural network has fine features in self-learning, self-organizing, self-association and fault-tolerant aspects, so the use of neural network has some theoretical value and practical value in achieving the technology of classifying and distinguishing cold-rolled steel sheet surface defects.First, this paper discusses the development and research status of the surface monitoring technology, and proposes a system design based on the performance requirements of cold-rolled steel sheet surface quality on-line monitoring system, builds and describes the overall system architecture and software process;Second, this dissertation has thoroughly researched on the technology of classifying and distinguishing cold-rolled steel sheet surface defects. By the research on feature extraction methods of geometry feature,gray feature and texture feature of the defect images, we design RBF network classifier based on the use of artificial neural network;Finally, we make a comparison between RBF network classifier and BP network classifier respectively from response time,the convergence number of iterations, a single iteration response time and recognition rate.We conclude RBF network classifier is better than BP network classifier,which indicates RBF network classifier has more application prospects and is closer to practical step.

Related Dissertations
More Dissertations