Special issue on machine learning in additive manufacturing

After 30 years of continuous development, additive manufacturing has successfully achieved mainstream acceptance as a popular manufacturing process. By creating products based on a 3D model, layer-by-layer, additive manufacturing allows for the production of complex parts with greater freedom in design optimization, surpassing traditional manufacturing techniques (Gibson et al. 2021). Meanwhile, machine learning has emerged as a hot technology with numerous applications in medical diagnosis, image processing, prediction, classification, learning association, and regression (Kotsiantis, Zaharakis, and Pintelas 2006). This has drawn increasing attention towards the use of machine learning in the manufacturing industry, particularly in the field of additive manufacturing (Jiang et al. 2020; Qin et al. 2022). The rapid progress of machine learning in additive manufacturing has brought us to this special issue, where we hope to bring together researchers with diverse research backgrounds in a common forum to accelerate the development of additive manufacturing technology through the aid of machine learning. We are enthusiastic about contributing to this cutting-edge research topic and driving further advancements in AM technology. In this special issue, there are nine papers accepted in the end with authors from Germany, India, Denmark, United States, Australia, Canada, Singapore, and China.

The paper ‘A survey of machine learning in additive manufacturing technologies’ (Jiang 2023) gives a state-of-the-art survey on machine learning in additive manufacturing and provides some guidelines for future applications of machine learning in additive manufacturing.

The next paper ‘Towards deep-learning-based image enhancement for optical camera-based monitoring system of laser powder bed fusion process’ (Zhang et al. 2022) introduces a super-resolution (SR) algorithm based on U-Net that can effectively enhance the details of optical camera monitoring images in the laser powder bed fusion (LPBF) process. A test setup was constructed in the laboratory to generate high-resolution images for training. To obtain accurate original images for validation purposes, low-resolution images were created by downscaling and blurring high-resolution images. The effectiveness of the SR algorithm was evaluated using the peak signal-to-noise ratio (PSNR) and plausibility of details as metrics. The results clearly demonstrate that this SR algorithm is capable of reconstructing intricate features from low-resolution images in the LPBF process.

The paper ‘Prediction of mechanical properties for acrylonitrile-butadiene-styrene parts manufactured by fused deposition modelling using artificial neural network and genetic algorithm’ (Mohd et al. 2022) uses a multi-parameter regression model to predict the mechanical properties of acrylonitrile-butadiene-styrene parts manufactured by fused deposition modelling. The model establishes a direct relation between choosing process parameters correctly and enhancing performance in fused deposition modelling. The model can provide optimal solutions for getting allotted output values.

The paper ‘Online monitoring for error detection in vat photopolymerization’ (Frumosu et al. 2023) proposes an online monitor system for bottom-up photopolymerization additive manufacturing systems. The data generated by the sensor is utilized to detect the detachment error of manufactured parts from the build platform, which cannot be physically observed by machine operators. If left undetected, this detachment can result in material waste and downtime, without stopping the ongoing build job. The online monitoring procedure is performed in two distinct phases: an offline training phase, followed by an online monitoring phase. During the offline phase, a prediction model is trained to be used in conjunction with a control chart for online monitoring. The monitoring control chart is particularly beneficial as it allows for the detection and recording of detachment predictions exclusively.

The next paper ‘Learning with supervised data for anomaly detection in smart manufacturing’ (Meiling et al. 2023) proposes a model selection architecture to automate the procedure of preprocessing input data and selecting the best combination of algorithms for anomaly detection. This architectural design plays an indispensable role in ensuring the production of superior quality products while enhancing quality control measures and business processes across a wide range of applications such as predictive maintenance and fault detection. Moreover, this framework is readily transferable to any smart manufacturing task within the domain of supervised learning.

The paper ‘Predictions of in-situ melt pool geometric signatures via machine learning techniques for laser metal deposition’ (Jiayu et al. 2022) propose a semi-dynamic and data-driven framework to ensure the fabricated quality in laser metal deposition. This framework is designed to correlate various process parameters, including laser power, scanning speed, powder feed rate, line energy density, and specific energy density, with features derived from melt pool thermal images such as melt pool width, area, mean temperature, and maximum temperature. The purpose of this correlation is to capture melt-pool-related features that are difficult to monitor directly, such as height, depth, and dilution. To develop and train machine learning models effectively, the authors conducted 60 single-track experiments to acquire sensing data and dimensions of the track cross-sections. Based on Spearman’s rank correlation coefficient, significant input features were selected for training these models.

The paper ‘Machine-learning-based monitoring and optimization of processing parameters in 3D printing’ (Tamir et al. 2022) proposes both open-loop and closed-loop machine learning models, which are integrated to monitor the impact of processing parameters on the quality of printed parts. Initially, an open-loop classification model is created by utilizing experimental 3D printing data to capture the correlation between processing parameters and printed part properties. Subsequently, a closed-loop control algorithm is developed, which combines the open-loop machine learning models with a fuzzy inference system to generate optimized processing parameters aimed at enhancing the quality of printed part properties.

The paper ‘Topology optimization of the vibrating structure for fused deposition modelling of parts considering a hybrid deposition path pattern’ (Guo, Ahmad, and Yongsheng 2022) introduces a novel concurrent topology optimization technique that aims to maximize the natural frequency of Fused Deposition Modelling parts, which are printed using a Hybrid Deposition Path (HDP) pattern. The proposed algorithm simultaneously optimizes both the structure’s shape and the raster directions of the substrate domain, utilizing the solid orthotropic materials with penalization (SOMP) method along with double layers of smoothing and projection (DSP).

The paper “Data-driven surrogate modelling of residual stresses in Laser Powder-Bed Fusion” (Lestandi et al. 2023) investigates using different machine learning models (i.e., multilayer perceptron (MLP), convolutional neural network (CNN) based on the U-Net architecture, and interpolation-based method based on mapping geometries onto a reference) for predicting the residual stress of parts fabricated by Laser Powder-Bed Fusion.

We would like to express our gratitude to the reviewers who provided valuable feedback and suggestions for improving the papers published in this special issue. Moreover, we extend our thanks to all the contributors who made the publication of this special issue possible. We would like to acknowledge the support of Professor Stephen T Newman, Editor-in-Chief of the International Journal of Computer Integrated Manufacturing, Professor Aydin Nassehi, Senior Editor, and Dr. Ray Zhong, Associate Editor, without which this publication would not have been possible. We sincerely hope this special issue will serve as a bridge between the academic community and practitioners, promoting the advancement of additive manufacturing using machine learning in the future.

Disclosure statement

No potential conflict of interest was reported by the author(s).