Interpretable machine learning-based rolling force prediction using multivariate industrial data during tandem cold rolling

Ironmaking & Steelmaking Processes Products and Applications, Год журнала: 2025, Номер unknown

Опубликована: Янв. 9, 2025

In the pursuit of intelligent manufacturing goals, industrial big data technology has emerged as a key enabler in advancing steel industry. Traditional rolling force (RF) models typically rely on from individual cold production lines, leading to lower accuracy and limited interpretability. To overcome this, an platform been developed, offering complete reliable dataset enhance performance RF prediction models. A data-driven machine learning framework is proposed, employing improved sparrow search algorithm optimise weighting parameters broad system. The Shapley additive explanations method further applied elucidate contributions multivariate features hot rolling, thereby enhancing interpretability predictions. proposed was validated line plant, demonstrating significant advantages over existing state-of-the-art Furthermore, this study demonstrates extensively elaborates impact predictive Industrial application validation that accurately predicts at head cold-rolled strip, enabling feedforward compensation for bending effectively improving flatness defects, confirming method's efficacy.

Язык: Английский

---

Audio-Based Engine Fault Diagnosis with Wavelet, Markov Blanket, ROCKET, and Optimized Machine Learning Classifiers

Sensors, Год журнала: 2024, Номер 24(22), С. 7316 - 7316

Опубликована: Ноя. 15, 2024

Engine fault diagnosis is a critical task in automotive aftermarket management. Developing appropriate fault-labeled datasets can be challenging due to nonlinearity variations and divergence feature distribution among different engine kinds or operating scenarios. To solve this task, study experimentally measures audio emission signals from compression ignition engines vehicles, simulating injector failures, intake hose absence of failures. Based on these faults, hybrid approach applied classify conditions that help the planning decision-making automobile industry. The proposed combines wavelet packet transform (WPT), Markov blanket selection, random convolutional kernel (ROCKET), tree-structured Parzen estimator (TPE) for hyperparameters tuning, ten machine learning (ML) classifiers, such as ridge regression, quadratic discriminant analysis (QDA), naive Bayes,

Язык: Английский

---

Enhancing Intermittent Spare Part Demand Forecasting: A Novel Ensemble Approach with Focal Loss and SMOTE

Logistics, Год журнала: 2025, Номер 9(1), С. 25 - 25

Опубликована: Фев. 8, 2025

Background: Accurate inventory management of intermittent spare parts requires precise demand forecasting. The sporadic and irregular nature demand, characterized by long intervals between occurrences, results in a significant data imbalance, where events are vastly outnumbered zero-demand periods. This challenge has been largely overlooked forecasting research for parts. Methods: proposed model incorporates the Synthetic Minority Oversampling Technique (SMOTE) to balance dataset uses focal loss enhance sensitivity deep learning models rare events. approach was empirically validated comparing model’s Mean Squared Error (MSE) performance Area Under Curve (AUC). Results: ensemble achieved 47% reduction MSE 32% increase AUC, demonstrating substantial improvements accuracy. Conclusions: findings highlight effectiveness method addressing imbalance improving prediction part providing valuable tool management.

Язык: Английский

---

Enhancing Defect Detection in Steel Plate Manufacturing with Explainable Machine Learning and SMOTE for Imbalanced Data

Journal of Materials Engineering and Performance, Год журнала: 2025, Номер unknown

Опубликована: Апрель 10, 2025

Язык: Английский

---

Industrial Big Data‐Driven Modeling and Prediction for Hot‐Rolled Strip Crown with Multigrade and Multispecification Data

steel research international, Год журнала: 2024, Номер 95(7)

Опубликована: Апрель 25, 2024

In the field of hot rolling big data, presence different steel types, specifications, and data heterogeneity poses significant challenges to accuracy stability using single machine learning regression technology for prediction. Therefore, this study proposes a hot‐rolled strip crown prediction method that combines clustering fusion modeling. First, article introduces relevant mechanism designing cluster strategies. The optimal strategy is determined through comparative experiments process parameters, size, main material components as features. Subsequently, K‐Means++ algorithm used effectively training testing datasets based on strategy, generating multiple clusters both datasets. Finally, establishes seven models match most suitable model each cluster, matching between rigorous testing. evaluation shows an R 2 value 0.829 root mean square error 3.974. experimental results show proposed outperforms traditional methods in solving multiclass classification heterogeneity, providing strong support intelligent control future.

Язык: Английский

---

LightGBM integration with modified data balancing and whale optimization algorithm for rock mass classification

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Окт. 3, 2024

Язык: Английский

---

Application of novel interpretable machine learning framework for strip flatness prediction during tandem cold rolling

Measurement, Год журнала: 2024, Номер unknown, С. 116516 - 116516

Опубликована: Дек. 1, 2024

Язык: Английский

---

Enhanced Online Strip Crown Prediction Model Based on KCGAN-ELM for Imbalanced Dataset

International Journal of Precision Engineering and Manufacturing, Год журнала: 2024, Номер 25(8), С. 1627 - 1637

Опубликована: Май 23, 2024

Язык: Английский

---

Detection, Prevention, and Monitoring Techniques for Industrial Equipment – a brief review

2022 14th International Conference on Electronics, Computers and Artificial Intelligence (ECAI), Год журнала: 2024, Номер unknown, С. 1 - 14

Опубликована: Июнь 27, 2024

Язык: Английский

---