Assessment of Outliers in Alloy Datasets Using Unsupervised Techniques
Jeffrey HawkJOM
Sign up for access to the world's latest research
checkGet notified about relevant papers
checkSave papers to use in your research
checkJoin the discussion with peers
checkTrack your impact
Abstract
Advancements in data analytics techniques have enabled complex, disparate datasets to be leveraged for alloy design. Identifying outliers in a dataset can reduce noise, identify erroneous and/or anomalous records, prevent overfitting, and improve model assessment and optimization. In this work, two alloy datasets (9-12% Cr ferritic-martensitic steels, and austenitic stainless steels) have been assessed for outliers using unsupervised techniques and supplemented with domain knowledge. Principal component analysis and k-means clustering were applied to the data, and points were assessed as outliers based on their distance from other points in the cluster and from other points in the dataset. The outlier characteristics were investigated to determine both cluster-specific and overall trends in the properties of the outlier points. The approach demonstrated here is extensible to other alloy datasets for outlier identification and evaluation to improve the reliability of machine learning and modeling predictions for advanced alloy design.
Related papers
Data Science Techniques, Assumptions, and Challenges in Alloy Clustering and Property Prediction
Journal of Materials Engineering and Performance, 2021
Data analytics methods have been increasingly applied to understanding materials chemistry, processing due to the manufacturing approach, and uni-axial and cyclic property relationships in the highly complex space of alloy design. There are several benefits to applying data analytics to this space, including the ability to manage non-linearities in the responses of the alloy attributes and the resulting mechanical properties. However, key difficulties in applying and understanding the results of data analytics include the often lack of reported assumptions and data processing steps necessary to improve interpretation and reproducibility in derived results. In this work, the methods used to generate clustering and correlation analyses for experimental 9% Cr ferritic-martensitic steel data were investigated and the resulting implications for mechanical property predictions were assessed. This work uses principal component analysis, partitioning around medoids, t-SNE, and k-means clustering to investigate trends in composition, processing and microstructure information with creep and tensile properties, building on work done previously using a smaller version of the same dataset. The initial assumptions, preprocessing steps and methods are investigated and outlined in order to depict the fine level of detail required to convey the steps taken to process data and produce analytical results. The variations in the resulting analyses are explored due to the influence of new and more varied data.
Anomaly detection for the prediction of ultimate tensile strength in iron casting production
2011
Mechanical properties are the attributes that measure the faculty of a metal to withstand several loads and tensions. In particular, ultimate tensile strength is the force a material can resist until it breaks. This property is one of the variables to control in the foundry process. The only way to examine this feature is to apply destructive inspections that make the casting invalid with the subsequent cost increment. Modelling the foundry process as an expert knowledge cloud allows machine-learning algorithms to forecast the value of a certain variable; in this case, the probability of a certain value of ultimate tensile strength for a foundry casting. Nevertheless, this approach needs to label every instance in the training dataset for generating the model that can foresee the value of ultimate tensile strength. In this paper, we present a new approach for detecting castings with an invalid ultimate tensile strength value based on anomaly detection methods. This approach represents correct castings as feature vectors of information extracted from the foundry process. A casting is then classified as correct or not correct by measuring its deviation to the representation of normality (correct castings). We show that this method is able to reduce the cost and the time of the tests currently used in foundries.
Evaluating data-driven algorithms for predicting mechanical properties with small datasets: A case study on gear steel hardenability
International Journal of Minerals, Metallurgy and Materials
Data-driven algorithms for predicting mechanical properties with small datasets are evaluated in a case study on gear steel hardenability. The limitations of current data-driven algorithms and empirical models are identified. Challenges in analysing small datasets are discussed, and solution is proposed to handle small datasets with multiple variables. Gaussian methods in combination with novel predictive algorithms are utilized to overcome the challenges in analysing gear steel hardenability data and to gain insight into alloying elements interaction and structure homogeneity. The gained fundamental knowledge integrated with machine learning is shown to be superior to the empirical equations in predicting hardenability. Metallurgical-property relationships between chemistry, sample size, and hardness are predicted via two optimized machine learning algorithms: neural networks (NNs) and extreme gradient boosting (XGboost). A comparison is drawn between all algorithms, evaluating the...
A data-driven machine learning approach to predicting stacking faulting energy in austenitic steels
Journal of Materials Science, 2017
The Material Genome Initiative (MGI) calls for establishing frameworks and adopting methodologies to accelerate materials discovery and deployment. The Integrated Computational Materials Engineering (ICME) approach and Materials Informatics leveraging materials data are two very important pillars to the initiative. This research is a data driven materials informatics approach to enable an ICME project on steel alloy design. For the alloy design problem there was a need to predict Stacking Fault Energy (SFE) for any untested alloy composition. SFE is a crucial parameter in determining different deformation regimes in austenitic steels. The SFE itself is dependent on the chemical composition and temperature in steels. There has been considerable study on determination of SFE in steels by experimental and computational methods. While the experimental methods investigate an alloy to find SFE, computational models have been constructed to predict SFE for a given composition and temperature. However, it is shown in this thesis that there are large inconsistencies in experimental data, as well as unavailability of robust computational models to predict SFE in truly multicomponent steel alloys. In this work, a data-driven machine learning approach to mine the literature of SFE in steels with the final aim of predicting deformation regimes for potentially unknown and untested alloy compositions has been demonstrated. Algorithms at the forefront of Machine Learning have been used to visualize the SFE data and then construct classifiers to predict SFE regime in steels. This machine-learning modeling approach can help accelerate alloy discovery of austenitic steels by linking composition to desired
Data-Driven Prediction of Mechanical Properties in Support of Rapid Certification of Additively Manufactured Alloys
Computer Modeling in Engineering & Sciences
Predicting the mechanical properties of additively manufactured parts is often a tedious process, requiring the integration of multiple stand-alone and expensive simulations. Furthermore, as properties are highly location-dependent due to repeated heating and cooling cycles, the properties prediction models must be run for multiple locations before the part-level performance can be analyzed for certification, compounding the computational expense. This work has proposed a rapid prediction framework that replaces the physics-based mechanistic models with Gaussian process metamodels, a type of machine learning model for statistical inference with limited data. The metamodels can predict the varying properties within an entire part in a fraction of the time while providing uncertainty quantification. The framework was demonstrated with the prediction of the tensile yield strength of Ferrium ® PH48S maraging stainless steel fabricated by additive manufacturing. Impressive agreement was found between the metamodels and the mechanistic models, and the computation was dramatically decreased from hours of physics-based simulations to less than a second with metamodels. This method can be extended to predict various materials properties in different alloy systems whose processstructure-property-performance interrelationships are linked by mechanistic models. It is powerful for rapidly identifying the spatial properties of a part with compositional and processing parameter variations, and can support part certification by providing a fast interface between materials models and part-level thermal and performance simulations.
Materials informatics for the screening of multi-principal elements and high-entropy alloys
Nature Communications, 2019
The field of multi-principal element or (single-phase) high-entropy (HE) alloys has recently seen exponential growth as these systems represent a paradigm shift in alloy development, in some cases exhibiting unexpected structures and superior mechanical properties. However, the identification of promising HE alloys presents a daunting challenge given the associated vastness of the chemistry/composition space. We describe here a supervised learning strategy for the efficient screening of HE alloys that combines two complementary tools, namely: (1) a multiple regression analysis and its generalization, a canonical-correlation analysis (CCA) and (2) a genetic algorithm (GA) with a CCA-inspired fitness function. These tools permit the identification of promising multi-principal element alloys. We implement this procedure using a database for which mechanical property information exists and highlight new alloys having high hardnesses. Our methodology is validated by comparing predicted hardnesses with alloys fabricated by arc-melting, identifying alloys having very high measured hardnesses.
Data-Driven Intelligent Model for the Classification, Identification, and Determination of Data Clusters and Defect Location in a Welded Joint
Processes
In this paper, a data-driven approach that is based on the k-mean clustering and local outlier factor (LOF) algorithm has been proposed and deployed for the management of non-destructive evaluation (NDE) in a welded joint. The k-mean clustering and LOF model algorithm, which was implemented for the classification, identification, and determination of data clusters and defect location in the welded joint datasets, were trained and validated such that three (3) different clusters and noise points were obtained. The noise points, which are regarded as the welded joint defects/flaws, allow for the determination of the cluster size, heterogeneity, and silhouette score of the welded joint data. Similarly, the LOF model algorithm was implemented for the detection, visualization, and management of flaws due to internal cracks, porosity, fusion, and penetration in the welded joint. It is believed that the management of welded joint flaws would aid the actualization of the Industry 4.0 concep...
Ijesrt International Journal of Engineering Sciences & Research Technology Relative Study of Outlier Detection Procedures
2016
Data Mining just alludes to the extraction of exceptionally intriguing patterns of the data from the monstrous data sets. Outlier detection is one of the imperative parts of data mining which Rexall discovers the perceptions that are going amiss from the normal expected conduct. Outlier detection and investigation is once in a while known as Outlier mining. In this paper, we have attempted to give the expansive and a far reaching literature survey of Outliers and Outlier detection procedures under one rooftop, to clarify the lavishness and multifaceted nature connected with each Outlier detection technique. Besides, we have likewise given a wide correlation of the different strategies for the diverse Outlier techniques. Outliers are the focuses which are unique in relation to or conflicting with whatever is left of the information. They can be novel, new, irregular, strange or uproarious data. Outliers are in some cases more fascinating than most of the information. The principle di...
Exploration of data science techniques to predict fatigue strength of steel from composition and processing parameters
Integrating Materials and Manufacturing Innovation, 2014
This paper describes the use of data analytics tools for predicting the fatigue strength of steels. Several physics-based as well as data-driven approaches have been used to arrive at correlations between various properties of alloys and their compositions and manufacturing process parameters. Data-driven approaches are of significant interest to materials engineers especially in arriving at extreme value properties such as cyclic fatigue, where the current state-of-the-art physics based models have severe limitations. Unfortunately, there is limited amount of documented success in these efforts. In this paper, we explore the application of different data science techniques, including feature selection and predictive modeling, to the fatigue properties of steels, utilizing the data from the National Institute for Material Science (NIMS) public domain database, and present a systematic end-to-end framework for exploring materials informatics. Results demonstrate that several advanced data analytics techniques such as neural networks, decision trees, and multivariate polynomial regression can achieve significant improvement in the prediction accuracy over previous efforts, with R 2 values over 0.97. The results have successfully demonstrated the utility of such data mining tools for ranking the composition and process parameters in the order of their potential for predicting fatigue strength of steels, and actually develop predictive models for the same.
Related papers
Comparison of the Application of Different Machine Learning Outlier Detection Methods on Actual Chemical Process Plant Data
Proceedings of the International Conference on Industrial Engineering and Operations Management
methods before training the surrogate models can enhance the prediction accuracy of the machine learning approximation models.
Machine learning for design, phase transformation and mechanical properties of alloys
Progress in Materials Science, 2022
Machine learning is now applied in virtually every sphere of life for data analysis and interpretation. The main strengths of the method lie in the relative ease of the construction of its structures and its ability to model complex non-linear relationships and behaviours. While application of existing materials have enabled significant technological advancement there are still needs for novel materials that will enable even greater achievement at lower cost and higher effectiveness. The physics underlining the phenomena involved in materials processing and behaviour however still pose considerable challenge and yet require solving. Machine learning can facilitate the achievement of these new aspirations and desires by learning from existing knowledge and data to fill in gaps that have so far been intractable for various reasons including cost and time. This paper reviews the applications of machine learning to various aspects of materials design, processing, characterisation, and some aspects of fabrication and environmental impact evaluation.
Adding Value to Sensor Data of Civil Engineering Structures: Automatic Outlier Detection
2018 9th International Conference on Information, Intelligence, Systems and Applications (IISA), 2018
This paper discusses the problem of outlier detection in datasets generated by sensors installed in large civil engineering structures. Since outlier detection can be implemented after the acquisition process, it is fully independent of particular acquisition processes as well as it scales to new or updated sensors. It shows a method of using machine learning techniques to implement an automatic outlier detection procedure, demonstrating and evaluating the results in a real environment, following the Design Science Research Methodology. The proposed approach makes use of Manual Acquisition System measurements and combine them with a clustering algorithm (DBSCAN) and baseline methods (Multiple Linear Regression and thresholds based on standard deviation) to create a method that is able to identify and remove most of the outliers in the datasets used for demonstration and evaluation. This automatic procedure improves data quality having a direct impact on the decision processes with regard to structural safety.
Identifying flow defects in amorphous alloys using machine learning outlier detection methods
Scripta Materialia, 2020
Shear transformation zones (STZs) are widely believed to be the fundamental flow defects that dictate the plastic deformation of amorphous alloys. However, it has been a long-term challenge to characterize STZs and their evolutions by experimental methods due to transient nature. Here we first introduced a consistent, automated, robust method to identify STZs by linear based machine learning outlier detection algorithms. We exemplify these algorithms to identify the atoms of STZs in Cu64Zr36 metallic glass sys- tem, and verify this data-driven model with a physical based model. It is revealed that the average STZ size slightly increases with decreasing cooling rate.
A Nonparametric Outlier Detection for Effectively Discovering Top-N Outliers from Engineering Data
Lecture Notes in Computer Science, 2006
We present a novel resolution-based outlier notion and a nonparametric outlier-mining algorithm, which can efficiently identify top listed outliers from a wide variety of datasets. The algorithm generates reasonable outlier results by taking both local and global features of a dataset into consideration. Experiments are conducted using both synthetic datasets and a real life construction equipment dataset from a large building contractor. Comparison with the current outlier mining algorithms indicates that the proposed algorithm is more effective.
Modern data analytics approach to predict creep of high-temperature alloys
Acta Materialia, 2019
A breakthrough in alloy design often requires comprehensive understanding in complex multicomponent/multi-phase systems to generate novel material hypotheses. We introduce a modern data analytics workflow that leverages high-quality experimental data augmented with advanced features obtained from high-fidelity models. Herein, we use an example of a consistently-measured creep dataset of developmental high-temperature alloy combined with scientific alloy features populated from a high-throughput computational thermodynamic approach. Extensive correlation analyses provide ranking insights for most impactful alloy features for creep resistance, evaluated from a large set of candidate features suggested by domain experts. We also show that we can accurately train machine learning models by integrating high-ranking features obtained from correlation analyses. The demonstrated approach can be extended beyond incorporating thermodynamic features, with input from domain experts used to compile lists of features from other alloy physics, such as diffusion kinetics and microstructure evolution.
Prediction of part density in additively manufactured maraging steel with supervised machine learning using pyrometer data
Materials Today: Proceedings, 2022
Additive Manufacturing (AM) / 3D printing technology is a game-changing technology for developing new improved solutions of product innovation with smart manufacturing advancements. One of the major challenges of AM manufactured metallic parts using laser powder bed fusion (LPBF) is the quality prediction of the printed samples under varying process parameters. This paper focuses on predicting the part density from pyrometer-based data using machine learning (ML) models, including Linear Regression (LR), Random Forest (RF), K-Nearest Neighbor (KNN), Support Vector Regression (SVR), Extreme Gradient Boosting (XGB), and Artificial Neural Network (ANN). Different pre-processing methods like Butterworth filter and thresholding have been compared with the raw pyrometer data-based analysis. Time-domain-based statistical features including mean, standard deviation, root mean square, entropy etc. have been used as inputs to the ML models. The six ML models were trained with and without feature selection (FS) to predict the part density. Among the regression algorithms used in this study, the best performance metrics R2 of 0.85 and 0.86 were obtained by RF regression using raw and filtered data respectively, while thresholding reduced model performance. Analysis reveals that the combined effect of laser power and scanning speed most influences the quality of printed parts. A subsequent experiment with the new process parameters chosen based on the data analysis was able to print parts with improved quality, thereby confirming the validity of our ML framework.
Prediction of Dimensional Changes of Low-Cost Metal Material Extrusion Fabricated Parts Using Machine Learning Techniques
Metals
Additive manufacturing (AM) is a layer-by-layer manufacturing process. However, its broad adoption is still hindered by limited material options, different fabrication defects, and inconsistent part quality. Material extrusion (ME) is one of the most widely used AM technologies, and, hence, is adopted in this research. Low-cost metal ME is a new AM technology used to fabricate metal composite parts using sintered metal infused filament material. Since the involved materials and process are relatively new, there is a need to investigate the dimensional accuracy of ME fabricated metal parts for real-world applications. Each step of the manufacturing process, from the material extrusion to sintering, might significantly affect the dimensional accuracy. This research provides a comprehensive analysis of dimensional changes of metal samples fabricated by the ME and sintering process, using statistical and machine learning algorithms. Machine learning (ML) methods can be used to assist re...
Prediction of Mechanical Properties of Steel using Data Science Techniques
International journal of recent technology and engineering, 2019
Investigating the Material Properties of Nodular Cast Iron from a Data Mining Perspective
Metals
Cast iron is a very common and useful metal alloy, characterized by its high carbon content (>4%) in the allotropic state of graphite. The correct shape and distribution of graphite are essential for ensuring that the material has the right properties. The present investigation examines the metallurgical and mechanical characterization of a spheroidal (nodular) cast iron, an alloy that derives its name and its excellent properties from the presence of graphite as spheroidal nodules. Experimental data are detected and considered from a data mining perspective, with the scope to extract new and little-known information. Specifically, a machine learning toolkit (i.e., Orange Data Mining) is used as a means of permitting supervised learners/classifiers (such as neural networks, k-nearest neighbors, and many others) to understand related metallurgical and mechanical features. An accuracy rate of over 90% can be considered as representative of the method. Finally, interesting considera...
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.