Data-driven mechanical property prediction and optimization of hot rolled microalloyed steels
Bibliographic record
Abstract
The primary objective of this research was to employ data-driven techniques to predict and optimize the mechanical properties of microalloyed steels during the thin slab direct rolling process. The data for this study were sourced from Algoma Steel Inc., located in Ontario, Canada, and the work was divided into two main parts.In the first part, Deep Neural Network (DNN) models were developed to predict the mechanical properties, specifically Ultimate Tensile Strength (UTS) and Lower Yield Strength (LYS), of Nb-based microalloyed hot-rolled strips. To introduce explainability into the DNN models, Game theory-based SHapely Additive exPlanations (SHAP) were utilized. The SHAP values provided insights into the combined effects of chemical composition and thermomechanical processing parameters. The influence of chemical composition was corroborated by physical metallurgy theory, and correlations were established with the empirical relationship of the No-recrystallization temperature (Tnr) from existing literature. Additionally, the data were analyzed using SIMS Mean Flow Stress (MFS) against the inverse temperature, with comparisons across different gauges and compositions to support the model explanations and suggest underlying metallurgical mechanisms. This segment of the study highlighted significant opportunities for optimization of alloy composition, which led to the second part of the research.The second part aimed to develop a data-driven framework for alloy design, considering the processing schedules of the rolling mill. Initially, seven different supervised machine learning (ML) algorithms were employed to model UTS and % Elongation for V-based microalloyed steel. Global feature importance was derived from SHAP values for these models. Model-agnostic conformal predictions were implemented to quantify uncertainty, enhancing the reliability of predictions. Given the challenges of inverse design in industrial contexts—such as multiple objectives, non-unique solutions, and large search spaces—the problem was approached as a multi-objective optimization (MOO) task focusing on the trade-off between strength and ductility, i.e. generating the best combination of strength and ductility. The best performing ML models for UTS and % Elongation were utilized as objective functions in the MOO, with the Non-dominated Sorting Genetic Algorithm II (NSGA-II) employed to derive optimized Pareto front solutions. The thermomechanical processing parameters were integrated as strict constraints in the decision variable space of the NSGA-II. To visualize the solutions, t-distributed Stochastic Neighbor Embedding (t-SNE) was used to map them along with original rolling data into a two-dimensional space, which was then clustered using the K-means algorithm. Select representative solutions from each cluster were chosen to identify unique alloys. This research provides key applications in developing online property prediction tools, enhancing process understanding and aiding in both process control, and alloy design
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".