Application of generalized regression neural network optimized by fruit fly optimization algorithm for fracture toughness in a pearlitic steel

Qiao, Lei; Liu, Yong; Zhu, Jingchuan

doi:10.1016/j.engfracmech.2020.107105

Cited by 45 publications

(19 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…ML is being exploited as an effective tool to help find material correlations and exhibits good applicability in gaining hidden insights from existing data to produce reliable decisions. [25,26] Our previous researches have successfully reported an application of ML methods to predict the interlamellar spacing, [27] fracture toughness, [28] and even construction of the constitutive equation. [29] Across these case studies, we identify the advantage of the ML approach in target property prediction.…”

Section: Introductionmentioning

confidence: 99%

Modeling of Alloying Effect on Isothermal Transformation: A Case Study for Pearlitic Steel

2021

Self Cite

View full text Add to dashboard Cite

Machine learning (ML) has been rapidly revolutionizing many fields and plays a vital role in materials science. To understand the microstructure of steels and even obtain optimized processing parameters, accurate predictions on the time–temperature–transformation (TTT) diagram are in urgent need. Herein, artificial neural network (ANN), support vector machine (SVM), and extreme learning machine (ELM) models are developed and compared for quantitatively predicting typical features, such as the nose tip temperature and incubation period, of TTT diagrams for pearlitic steels. As an important evaluating indicators of prediction accuracy, the correlation coefficient (R2), the root mean square error (RMSE), and the mean absolute percent error (MAPE) are used to evaluate the proposed models. The results show that the proposed SVM model exhibits higher prediction accuracy compared with the ANN and ELM and demonstrates the advantage of the machine learning approach in predictions of TTT diagrams. Moreover, a series of experiments demonstrate that the proposed SVM predictor performs favorably, with satisfactory prediction accuracy. The microstructures of the isothermal transformation products of the prepared pearlitic steel are investigated using dilatation and microscopy technique.

show abstract

Section: Introductionmentioning

confidence: 99%

Modeling of Alloying Effect on Isothermal Transformation: A Case Study for Pearlitic Steel

2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…With recent developments, machine learning (ML) approaches are becoming potential assistants for constructing models and discovering correlations to predict material characteristics, including compositions, microstructures, and properties. [ 18–27 ] For example, Shi et al [ 23,24 ] review the application of the ML method in materials science and emphasize their advances to predict independent properties and accelerate development of the industrial chain. As known, accurate prediction of the volume fraction of resulting phase with a given combination of alloying elements has become increasingly important to the development and applications of new steels.…”

Section: Introductionmentioning

confidence: 99%

Phase Prediction of High Carbon Pearlitic Steel: An Improved Model Combining Mind Evolutionary Algorithm and Neural Networks

et al. 2021

Self Cite

View full text Add to dashboard Cite

Accurate phase constitution prediction is crucial for guiding the new steel design with desirable properties. This article uses three machine learning (ML) algorithms, backpropagation neural network (BPNN), radial basis function neural network (RBFNN), and fuzzy neural network (FNN), to compare the accuracy of predictive model. Toward the collected data, statistical measure of correlation is taken in the present work by determining Pearson correlation coefficient (PCC). The results show that the testing accuracy using BPNN is higher than the corresponding testing accuracy using RBFNN and FNN. With these understandings, the well‐known optimization algorithms, namely, mind evolutionary algorithm (MEA), are implemented to find optimal hyperparameters set that is able to induce a higher predictive capability. The resulting MEA‐BPNN further enhances the prediction results in the present dataset and efficiently explores the relationship between alloying elements and phase constitution. Finally, the reliability and practicability of the model are verified by experimental measurement on the ferrite content for prepared steels, and their mechanical properties are tested for engineering applications. As such, the work proposes a useful MEA‐BPNN model for predicting the phases of high carbon pearlite steel and provides an alternative route of designing new steels in a given system.

show abstract

“…Situations where only small experimental datasets are available may be encountered in many fields because generating large datasets is time-and-cost prohibitive [14,27,35,36]. Therefore, different methods have been developed and implemented in ANN modeling to cope with the small dataset conditions.…”

Section: Introductionmentioning

confidence: 99%

“…Therefore, different methods have been developed and implemented in ANN modeling to cope with the small dataset conditions. Examples of these methods include using simulated data [37], using virtual data [38], using multiple runs for models development and surrogate data analysis for model validation [39], using duplicated experimental runs [9], using stacked auto-encoder pre-training [14], using analytical models with errors revised by intelligent algorithms [35], using optimization aided generalized regression approach [36], and simultaneously considering data samples with their posterior probabilities [40]. Candelieri et al [37] used datasets obtained from finite element simulations to develop an ANN model for diagnosing and predicting cracks in aircraft fuselage panels.…”

Section: Introductionmentioning

confidence: 99%

“…The results showed that the Arrhenius-type model with errors revised by ANN gives the best prediction accuracy among all other models, even with a small training dataset. Qiao et al [36] proposed the use of a generalized regression neural network (GRNN) optimized by fruit-fly optimization algorithm (FOA) to improve the predictive accuracy when using small training datasets. Using a small dataset comprising only 14 data points, the proposed approach was used for modeling the relationship between alloying elements and fracture toughness of steel alloys.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Novel Assisted Artificial Neural Network Modeling Approach for Improved Accuracy Using Small Datasets: Application in Residual Strength Evaluation of Panels with Multiple Site Damage Cracks

2020

View full text Add to dashboard Cite

An artificial neural network (ANN) extracts knowledge from a training dataset and uses this acquired knowledge to forecast outputs for any new set of inputs. When the input/output relations are complex and highly non-linear, the ANN needs a relatively large training dataset (hundreds of data points) to capture these relations adequately. This paper introduces a novel assisted-ANN modeling approach that enables the development of ANNs using small datasets, while maintaining high prediction accuracy. This approach uses parameters that are obtained using the known input/output relations (partial or full relations). These so called assistance parameters are included as ANN inputs in addition to the traditional direct independent inputs. The proposed assisted approach is applied for predicting the residual strength of panels with multiple site damage (MSD) cracks. Different assistance levels (four levels) and different training dataset sizes (from 75 down to 22 data points) are investigated, and the results are compared to the traditional approach. The results show that the assisted approach helps in achieving high predictions’ accuracy (<3% average error). The relative accuracy improvement is higher (up to 46%) for ANN learning algorithms that give lower prediction accuracy. Also, the relative accuracy improvement becomes more significant (up to 38%) for smaller dataset sizes.

show abstract

Application of generalized regression neural network optimized by fruit fly optimization algorithm for fracture toughness in a pearlitic steel

Cited by 45 publications

References 28 publications

Modeling of Alloying Effect on Isothermal Transformation: A Case Study for Pearlitic Steel

Modeling of Alloying Effect on Isothermal Transformation: A Case Study for Pearlitic Steel

Phase Prediction of High Carbon Pearlitic Steel: An Improved Model Combining Mind Evolutionary Algorithm and Neural Networks

A Novel Assisted Artificial Neural Network Modeling Approach for Improved Accuracy Using Small Datasets: Application in Residual Strength Evaluation of Panels with Multiple Site Damage Cracks

Contact Info

Product

Resources

About