Prediction of Endpoint Sulfur Content in KR Desulfurization Based on the Hybrid Algorithm Combining Artificial Neural Network With SAPSO

Cao

et al. 2020

IEEE Access

Self Cite

In the present work, to solve the problem of the lacking enough labeled training data for deep learning, a safe semi-supervised regression supporting Bayesian optimization deep neural network (SAFER-BODNN) model was proposed to establish mechanical property prediction model of hot-rolled strips. The Pearson correlation coefficient was applied to reduce the data dimension. The safe semi-supervised regression was implemented to add the pseudo labels to the unlabeled data for training dataset expansion. The deep neural network was trained with Bayesian optimization to determine the optimal hyper-parameters of the network. The results show that the SAFER-BODNN model achieves good accuracy for mechanical property prediction of hot-rolled strips with R of 0.9610 for yield strength, 0.9682 for tensile strength, and 0.8619 for elongation, respectively. Compared with the deep neural network trained on the labeled dataset, the SAFER-BODNN model obtains stable smaller predicted errors. Among all the variables, C content and Mn content have large influence on the yield strength and tensile strength, coiling temperature has the largest influence on the elongation. The investigation makes full use of unlabeled data to elevate the accuracy of the deep neural network, and also provides a way for deep learning modeling when the data are insufficient.

“…In order to quantify the degree of deviation between the predicted values and the actual values, the RMSE and MAE were applied, which can be calculated by Eqs. (11) and (12) [44,45].…”

Section: Resultsmentioning

confidence: 99%

Elevating Prediction Performance for Mechanical Properties of Hot-Rolled Strips by Using Semi-Supervised Regression and Deep Learning

Cao

et al. 2020

IEEE Access

Self Cite

“…Since the model has been updated with the new regressor, the next iteration will take place until the model complete the training. The boosting process can be expressed as Equation ( 8)- (11). [14] f 0 ðxÞ ¼ arg min f ðxÞ Δf t ðxÞ ¼ ρ t h t (10)…”

Section: Machine Learning Modelsmentioning

confidence: 99%

“…To eliminate outliers, the Pauta criterion is employed. [ 11,12 ] Assuming x = { x 1 , x 2 , x 3 , …, x n , …, x N } is the experimental data set. The average of

\overset{x}{true}

, the residual error of

r_{n}

, and the standard deviation of σ are calculated by Equation (1)–(3), respectively.

\overset{false¯}{x} = \frac{1}{N} {false\sum}_{n = 1}^{N} x_{n}

r_{n} = x_{n} - \overset{false¯}{x}

σ = \sqrt{\frac{1}{N} {false\sum}_{1}^{N} {false(x_{n} - \overset{x}{true} false)}^{2}}

where N is the number of samples.…”

Section: Data Preprocessingmentioning

confidence: 99%

Comparison of the Prediction of BOF End‐Point Phosphorus Content Among Machine Learning Models and Metallurgical Mechanism Model

Zhang

et al. 2022

steel research int.

Self Cite

As a harmful element, phosphorus will cause cold brittleness that decreases the strength of steel. In order to ensure the mechanical properties of steel products, the phosphorus is generally removed from molten steel in BOF (basic oxygen furnace) steelmaking process. Therefore, the prediction of end-point P content in BOF is of great significance for steelmaking production.Several methods are usually employed to predict the end-point P content. The first one is using simple empirical formulas based on the previous production data. Through this method, the steelmaking workers can quickly predict the end-point P content on site and then promptly adjust the process parameters of BOF. However, the accuracy of this method is pretty low, and the prediction can only be used as a reference. Another method is the end-point P content prediction with the metallurgical mechanism from the perspective of dephosphorization kinetics. The dephosphorization kinetic model is established based on the coupled reaction model or the first principle. [1] The advantage of this method is that one can obtain not only the end-point P content but also the P content during the process. Although the metallurgical mechanism model (MMM) is with high accuracy, it takes as long as 20 min to complete the calculation for one heat, which cannot satisfy the quick steelmaking process. [2] With the progresses of computer equipment and artificial intelligence algorithms, the machine learning shows its outstanding capabilities in various fields in classification, regression, data mining, image recognition, and so on. The machine learning has also been applied in BOF end-point P content prediction for many years. He et al. proposed a prediction model based on the principal component analysis (PCA) and back propagation neural network (BPNN). [3] The results showed that the PCA-BPNN model had the highest prediction accuracy compared with the multiple linear regression (MLR) model and the normal BPNN model. Phull et al. and Li et al. used twin support vector machines (TWSVM) to classify the degrees of phosphorus removal (such as "High" and 'Low"). [4,5] The models achieved a high accuracy of 99.9% in classification, and the running time was shorter than the other models such as logistic regression and SVM radial basis. Wang et al. established a hybrid method for the end-point prediction of BOF. [6] This model could not only predict the end-point P content but also forecast the end-point C content and temperature. After applying the hybrid algorithm to improve the theoretical model and artificial neural networks (ANN), the performances of the models were improved. Moreover, many researchers applied different models to predict the end-point P content, such as membrane algorithm evolving extreme learning machine, [7] group method of data handling polynomial neural network, [8] and monotone-constrained BP neural network. [9]

“…The three indexes, correlation coefficient (R), root mean square error, and mean absolute relative error are all better than MLR. [ 19 ] Support vector regression (SVR) as a regression method based on binary classification model, can realize data regression or solution through margin maximization in feature space. Model for predicting steel production volume was constructed based on this method.…”

Section: Literature Reviewmentioning

confidence: 99%

An Integration Model for Converter Molten Steel End Temperature Prediction Based on Bayesian Formula

Feng

et al. 2021

steel research int.

Highly accurate prediction of converter molten steel end temperature is important foundation of realizing intelligent smelting in converter procedure. Single model prediction of converter end temperature has problems such as weak generalization ability and difficulty in increasing accuracy. To tackle these problems, this review proposes a modeling method based on Bayesian formula that dynamically integrate multiple models with different features. First, various data patterns and features are explored using different modeling methods and respective prediction model for converter molten steel end temperature is constructed. Then, the value range of the objective is discretized and the confidence levels of the prediction results from different models are computed based on Bayesian formula. Last, the prediction values of different models are weighted according to their confidence level and a comprehensive prediction result is obtained. Testing and comparison are carried out on the integration model and the three single models (support vector regression, random forest, and BP neural network) using actual production data. Results show that the integration model based on Bayesian formula can further increase the prediction accuracy effectively on top of single model prediction accuracy.