A Capillary Quality Prediction Method Based on Improved Elm Algorithm

Abstract

The invention relates to a capillary quality prediction method based on an improved ELM algorithm. The steps are: collecting multiple groups of historical field data in the capillary perforation process to construct a training set; determining the input layer, output layer and hidden layer of the integrated ELM network according to the collected field data Containing layers; combined with multiple commonly used excitation functions, by setting weights, the excitation function of the integrated ELM network is determined; the genetic algorithm is used to optimize each weight in the excitation function of the integrated ELM network to obtain the optimal excitation function; using training The integrated ELM network is trained to complete the construction of the integrated ELM network; the data in actual production is input into each sub-network of the integrated ELM network, and the output results of each sub-network are obtained, and then the output forecast results of the integrated ELM network are obtained. That is, the prediction result of capillary quality. The invention inherits the fast performance of the ELM model and the robustness of the integration method, and can more accurately predict the quality of the capillary.

Description

technical field [0001] The invention belongs to the quality prediction technology in the field of regression technology, and in particular relates to a capillary quality prediction method based on an improved ELM algorithm. Background technique [0002] As the first process of seamless steel pipe production, piercing has a very important impact on the quality of steel pipes; the quality problems caused by the piercing process will not be alleviated in the subsequent process, but will cause more serious quality problems of steel pipes; so , the establishment of a capillary quality prediction model using the collected perforation process data has very important guiding significance for the steel rolling process; common prediction methods are mainly based on time series method, Kalman filter method, neural network, support vector machine (SVM), etc.; The prediction results of the time series method are unstable, and its model parameters are difficult to determine; the neural ne...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the shortcomings of the capillary quality prediction method in the prior art that the sample data has noise interference and the prediction result of the ELM model is unstable, the present invention proposes a capillary quality prediction method based on the improved ELM algorithm to achieve the purpose of improving the accuracy of capillary prediction

Images