Computer-simulated experiments have become a cost effective way for engineers to replace real experiments in the area of product development. However, one single computer-simulated experiment can still take a significant amount of time. Hence, in order to minimize the amount of simulations needed to investigate a certain design space, different approaches within the design of experiments area are used. One of the used approaches is to minimize the time consumption and simulations for design space exploration through response surface modeling. The traditional methods used for this purpose are linear regression, quadratic curve fitting and support vector machines. This paper analyses and compares the performance of four machine learning methods for the regression problem of response surface modeling. The four methods are linear regression, support vector machines, M5P and random forests. Experiments are conducted to compare the performance of tree models (M5P and random forests) with the performance of non-tree models (support vector machines and linear regression) on data that is typical for concept evaluation within the aerospace industry. The main finding is that comprehensible models (the tree models) perform at least as well as or better than traditional black-box models (the non-tree models). The first observation of this study is that engineers understand the functional behavior, and the relationship between inputs and outputs, for the concept selection tasks by using comprehensible models. The second observation is that engineers can also increase their knowledge about design concepts, and they can reduce the time for planning and conducting future experiments.

In engineering, design analyses of complex products rely on computer simulated experiments. However, high-fidelity simulations can take significant time to compute. It is impractical to explore design space by only conducting simulations because of time constraints. Hence, surrogate modelling is used to approximate the original simulations. Since simulations are expensive to conduct, generally, the sample size is limited in aerospace engineering applications. This limited sample size, and also non-linearity and high dimensionality of data make it difficult to generate accurate and robust surrogate models. The aim of this paper is to explore the applicability of Random Forests (RF) to construct surrogate models to support design space exploration. RF generates meta-models or ensembles of decision trees, and it is capable of fitting highly non-linear data given quite small samples. To investigate the applicability of RF, this paper presents an approach to construct surrogate models using RF. This approach includes hyperparameter tuning to improve the performance of the RF's model, to extract design parameters' importance and \textit{if-then} rules from the RF's models for better understanding of design space. To demonstrate the approach using RF, quantitative experiments are conducted with datasets of Turbine Rear Structure use-case from an aerospace industry and results are presented.

The design of aircraft engines involves computationally expensive engineering simulations. One way to solve this problem is the use of response surface models to approximate the high-fidelity time-consuming simulations while reducing computational time. For a robust design, sensitivity analysis based on these models allows for the efficient study of uncertain variables’ effect on system performance. The aim of this study is to support sensitivity analysis for a robust design in aerospace engineering. For this, an approach is presented in which random forests (RF) and multivariate adaptive regression splines (MARS) are explored to handle linear and non-linear response types for response surface modelling. Quantitative experiments are conducted to evaluate the predictive performance of these methods with Turbine Rear Structure (a component of aircraft) case study datasets for response surface modelling. Furthermore, to test these models’ applicability to perform sensitivity analysis, experiments are conducted using mathematical test problems (linear and non-linear functions) and their results are presented. From the experimental investigations, it appears that RF fits better on non-linear functions compared with MARS, whereas MARS fits well on linear functions.

One of the crucial aspects of additive manufacturing is the monitoring of the welding process for quality assurance of components. A common way to analyse the welding process is through visual inspection of melt-pool images to identify possible defects in manufacturing. Recent literature studies showed the potential use of prediction models for defects classification to speed up the manual verification criteria since a huge data is generated from the additive manufacturing. Although a huge image data is available, the data needs to be labelled manually by experts which results in small sample datasets. Hence, to model small sample sizes and also to acquire the importance of parameters, we opted a traditional machine learning method, Random Forests (RF). For feature extraction, we opted for the Polar Transformation to explore its applicability using the melt-pool image dataset and a publicly available shape image dataset. The results show that RF models with Polar Transformation performed the best on our case study datasets and the second-best for the public dataset when compared to the Histogram of Oriented Gradients, HARALICK, XY-projections of an image, and Local Binary Patterns methods. As such, the Polar Transformation can be considered as a suitable compact shape descriptor. © 2020 IEEE.

Large amount of data are generated from in-situ monitoring of additive manufacturing (AM) processes which is later used in prediction modelling for defect classification to speed up quality inspection of products. A high volume of this process data is defect-free (majority class) and a lower volume of this data has defects (minority class) which result in the class-imbalance issue. Using imbalanced datasets, classifiers often provide sub-optimal classification results i.e. better performance on the majority class than the minority class. However, it is important for process engineers that models classify defects more accurately than the class with no defects since this is crucial for quality inspection. Hence, we address the class-imbalance issue in manufacturing process data to support in-situ quality control of additive manufactured components.  For this, we propose cluster-based adaptive data augmentation (CADA) for oversampling to address the class-imbalance problem. Quantitative experiments are conducted to evaluate the performance of the proposed method and to compare with other selected oversampling methods using AM datasets from an aerospace industry and a publicly available casting manufacturing dataset. The results show that CADA outperformed random oversampling and the SMOTE method and is similar to random data augmentation and cluster-based oversampling. Furthermore, the results of the statistical significance test show that there is a significant difference between the studied methods.  As such, the CADA method can be considered as an alternative method for oversampling to improve the performance of models on the minority class. 

This paper aims to address data labelling issues in process data to support in-situ process monitoring of additive manufactured components. For this, we adopted an active learning (AL) approach to minimise the manual effort for data labelling for classification models. In this study, we present an approach that utilises pre-trained models to extract deep features from images, and clustering and query by committee sampling to select the representative samples to build defect classification models. We conduct quantitative experiments to evaluate the proposed method's performance and compare it with other selected state-of-the-art AL approaches using a dataset of additive manufacturing (AM) and a publicly available dataset. The experimental results show that the proposed approach outperforms AL with committee based sampling, and AL with clustering and random sampling. The results of the statistical significance test show that there is a significant difference between the studied AL approaches. Hence, the proposed AL approach can be considered an alternative method to reduce labelling costs when building defects classification models, whose generalizability is most likely plausible.