Jul 2026· Ironmaking & Steelmaking: Processes, Products and Applications· 0 citations· 32 references
TL;DR
A prediction method based on particle swarm optimisation (PSO) and a one-dimensional convolutional neural network (1D-CNN) was developed and integrated into an online system that enables single tube wall thickness prediction, sawing parameter calculation, and batch visualisation for process adjustment and sawing decisions.
Abstract
To improve the accuracy of axial wall thickness prediction and reduce reliance on manual measurements in hot-rolled steel tube production, a prediction method based on particle swarm optimisation (PSO) and a one-dimensional convolutional neural network (1D-CNN) was developed. A dataset was constructed from 131 industrial samples collected from the production line. An input variable set was established to characterise entry wall thickness, geometric and overall deformation indicators, pass schedule profile features, speed schedule profile and thermal conditions. PSO was then employed to optimise the key hyperparameters of the 1D-CNN. The proposed model was compared with several machine learning models. The results showed that the PSO-1D-CNN model achieved the best predictive performance, with a test root mean square error of 0.0281, a mean absolute error of 0.0217, and a coefficient of determination
R
2
of 0.913. Further interpretability analysis using SHapley Additive exPlanations revealed that the centroid of the reduction distribution, the number of stands, the diameter-to-wall ratio, the tube temperature at the sizing exit, and the radial compression ratio were the most influential variables affecting the predictions. Finally, the proposed model was integrated into an online system. This system enables single tube wall thickness prediction, sawing parameter calculation, and batch visualisation for process adjustment and sawing decisions.
The potential effectiveness of the proposed robustness-oriented evaluation framework for ML-assisted Ra prediction under limited-data machining conditions is supported, and ELM achieved the highest prediction accuracy.
Phong Thanh Huynh, T. Nguyen, Huynh Thanh Thuong· Engineering Research Express· 0 citations
Coiling temperature is a pivotal process parameter for hot-rolled strip steel, which directly determines the microstructure and mechanical properties of final products. Affected by the coupling of multiple process variables, coiling temperature presents strong nonlinearity and complex time-varying characteristics. Traditional heat transfer mechanism models, Random Forest (RF), Extreme Learning Machine (ELM) and single Long Short-Term Memory (LSTM) networks fail to fully explore the deep correlation among variables. In addition, their hyperparameters are generally selected by manual trial-and-error, leading to unsatisfactory prediction accuracy and poor robustness in practical production. To address the above limitations, this paper proposes a novel prediction model named WOA-CNN-GRU-SE, where the Whale Optimization Algorithm (WOA) is adopted for parameter optimization. Firstly, Convolutional Neural Network (CNN) is utilized to extract local coupling features from various working condition parameters. Secondly, the Squeeze-and-Excitation (SE) attention mechanism is applied to adaptively recalibrate channel weights, which enhances key features closely related to temperature variation and suppresses redundant interference information. Afterwards, Gated Recurrent Unit (GRU) is employed to conduct in-depth learning of temporal features. Furthermore, WOA is used to globally optimize critical hyperparameters, including learning rate, the number of GRU hidden units and L2 regularization coefficient, so as to eliminate the drawbacks of manual parameter tuning. Comparative experiments are conducted on actual production data from a hot rolling line. The results demonstrate that the proposed model outperforms CNN-GRU, CNN-GRU-SE, LSTM, RF and ELM in prediction performance. Its hit rate reaches 92.56% within the industrial error range of ±6 °C. This model effectively realizes accurate prediction of coiling temperature under complex working conditions and possesses great application potential in industrial practice.
Precise prediction and control of porosity in laser powder bed fusion (L-PBF) directly enhances the performance of additively manufactured components. This study addresses the need for comprehensive machine learning (ML) analysis of pore characteristics through comparative evaluation of multiple ML models based on accuracy and reliability. Five supervised ML models-linear regression (LR), Gaussian process regression, decision tree regression, artificial neural networks (ANNs), and random forest regression (RFR)-were utilized to predict pore characteristics in 316L stainless steel components fabricated via L-PBF. Model performance was systematically assessed using three key metrics: root mean square error, mean absolute error, and coefficient of determination (
R
2
). These metrics were calculated from process-condition averages under grouped cross-validation to ensure robust evaluation. Material characterization of specimens produced using pore-optimized printing parameters further validated the predictive accuracy of the models. Our results revealed that distinct models were optimal for different pore-related targets: LR outperformed others for porosity prediction, RFR excelled in estimating average pore diameter, and ANN delivered the highest accuracy for average pore roundness. Response surfaces generated from the optimal models delineated a processing window (laser power: 150–250 W; scan speed: 800–1200 mm/s; layer thickness: 0.04 mm; and hatch spacing: 0.09 mm) associated with minimized porosity and improved pore morphology. Notably, a significant inverse correlation was observed between predicted porosity and critical mechanical properties (including yield strength, tensile strength, and elongation at break), which further corroborated the practical utility of the proposed predictive workflow.
Shi-Yu Liu, Cheng Liu, Xiao Xue et al.· Metallurgical Research &...· 0 citations
The purpose of this paper is to develop machine learning (ML) models for prediction of surface roughness and cutting forces of 42CrMo4 steel in hard turning process.
A full factorial experimental design with four input parameters: cutting speed, depth of cut, feed and insert radius was used to develop ML models for predicting the performance of turning process. The backward linear regression, random forest (RF) and XGBoost were used. Also, for the linear regression model and for the best RF and XGBoost model five-fold cross validation was done to confirm that the models provide reliable generalization estimates rather than performance dependent on a single data split.
The XGBoost model demonstrates the most compact clustering of residuals with fewer large errors, indicating better overall stability and predictive consistency compared to the linear regression and RF models.
The application of different ML methods with monitoring of standardized residuals on unseen data confirms the reliability of the developed models in real application conditions.
This study provides a structured and comparative modeling framework across multiple output variables, where backward linear regression, RF and XGBoost models were developed. Several architectural and hyperparameter variations of the RF and XGBoost models were evaluated to ensure optimal configuration for each output. Also, variable influence was examined through permutation feature importance for ensemble models and statistical significance testing for linear regression, enabling interpretation and discussion of the influence of input variables on selected outputs.
Mirza Pašić, Aleksandar Živković, K. Muhamedagic et al.· Engineering computations· 0 citations
In the current context of the manufacturing industry, optimizing the cutting parameters to achieve a controlled surface roughness involves costly and time-consuming experimental efforts. The present study addresses this challenge by developing robust machine learning-based approximate functions for predicting surface roughness (Ra) resulting from toroidal milling on a five-axis CNC. The research includes an experimental design conducted under real production conditions on C45 steel. The relatively small experimental dataset was augmented, normalized, and then scripts were written for four prediction models: two artificial neural network architectures and two models based on decision trees. Their performance was analyzed based on MSE, RMSE, R2, and MRA metrics. The results obtained reveal significant differences between the models, highlighting solutions with high accuracy, excellent robustness, and superior generalization capacity for new data. The study highlights the high potential of prediction models in optimizing machining processes, providing an effective way to reduce costly physical experiments and increase productivity in industrial environments.
M. Banica, A. Osan, Andrei Filip· Machines· 0 citations
The paper presents an approach to constructing predictive models for the physical and mechanical properties of elastomeric composites using machine learning methods. The relevance of the study is driven by the need to accelerate the development of new materials and reduce the labor intensity of full-scale experiments. An automated machine learning algorithm is proposed, encompassing stages of input data unification, feature space formation, and comparative analysis of regression models (Random Forest, Gradient Boosted Decision Trees, Gradient Tree-Boosting Tweedie, Poisson Regression, Light Gradient Boosting Machine и Stochastic Dual Coordinate Ascent). During experimental validation on datasets containing formulation data with varying content of sulfur, natural rubber (NR), and zinc oxide, predictions were made for theoretical density, Karrer plasticity, brittleness temperature, and curing temperature. It was established that ensemble methods demonstrate the highest predictive capability; however, model accuracy significantly depends on sample representativeness. Intervals of the studied parameters (particularly the 140–150 °C range for curing temperature) characterized by increased prediction uncertainty were identified, requiring additional algorithm calibration. The obtained results confirm the effectiveness of the proposed approach for formulation optimization and identification of hidden dependencies in the “composition-property” system.
M. Maslova, V. Kablov, A. Rybanov· Computational nanotechnology· 0 citations