Coiling Temperature Prediction Study Of 2250mm Hot Strip Mill

Laminar cooling is an essential manner in hot rolling production,which can rapidly cool the strip to the anticipated coiling temperature,to correctly entire the coiling of strip products.The correct manage of strip coiling temperature is the most vital undertaking in hot rolling production.The improvement of coiling temperature prediction accuracy can reduce the labor cost and material loss to a certain extent,and also improve the product sheet performance and coiling rate.Accurate prediction of coiling temperature and establishment of high precision prediction model can not solely provide a new way of thinking for the actual production of hot rolling,however additionally is an important prerequisite for enhancing the performance of strip steel sheets.In this context,this thesis conducts an indepth study on strip coiling temperature prediction.Makes full use of the production data of Baosteel hot rolling line and combined with artificial intelligence to achieve high accuracy prediction of strip coiling temperature,and the main research contents are as follows.(1)By consulting the literature in related fields and analyzing the current status of hot rolling research,researching the hot rolling production site,collecting production data,using the actual data after pre-processing,determining the variables affecting the strip coiling temperature by combining the actual process mechanism and expert experience,and establishing the strip coiling temperature prediction model.(2)For the same group of hot rolling data,the neural network models based on BP,Elman and ELM were established to simulate the prediction,determine the optimal excitation function and the wide variety of nodes in the implicit layer of every model,and simulate the evaluation indexes such as root mean square error.The experimental consequences exhibit that the ELM model has higher prediction accuracy and is higher than the BP and Elman models,so the ELM model is used for this thesis.(3)To tackle the problems that ELM weights and thresholds are random and easy to lead to the discount of mannequin stability,two improved optimization algorithms are introduced to optimize its parameters.First,for the gray wolf optimization algorithm,Henon mapping,small-hole imaging strategy and weight aspect method are delivered for improvement,and it is located that the improved gray wolf algorithm is better than the contrast algorithm through test functions and statistical assessments,etc.,and the best result of finding the best result;second,for the Harris Hawk algorithm,it is proposed to use normal cloud,random inverse and dynamic perturbation strategies for improvement,and the consequences results show that the improved algorithm proposed in this thesis.The test results exhibit that the improved algorithm proposed in this thesis is better than the five basic algorithms and the three improved Harris Hawk algorithms compared,and the algorithm has the strongest performance in finding the best.(4)According to the distinct overall performance of the improved Gray Wolf algorithm and the improved Harris Hawk algorithm on the test function,the single steel prediction model based on IGWO-ELM and the multi-steel prediction model based on IHHO-ELM were established,in which the prediction temperature error of IGWO-ELM model was 91.1%and 96.7% at ±3°C and ±4°C,and the prediction temperature error of IHHO-ELM model was 86.7% and 92.7% at ±4°C and ±5°C respectively,and the prediction results of both models were better than the comparison models and much higher than the temperature requirements of the hot rolling line,which have good application prospects and guiding significance.