Deep Learning-Based Data Fusion Method for In Situ Porosity Detection in Laser-Based Additive Manufacturing

Abstract

Laser-Based Additive Manufacturing (LBAM) provides unrivalled design freedom with the ability to manufacture complicated for a wide range of engineering applications. Melt pool is one of the most important signatures in LBAM and is indicative of process anomalies and part defects. High-speed thermal images of the melt pool captured during LBAM make it possible for in-situ melt pool monitoring and porosity prediction. This paper aims to broaden current knowledge of the underlying relationship between process and porosity in LBAM and provide new possibilities for efficient and accurate porosity prediction. We present a deep learning-based data fusion method to predict porosity in LBAM parts by leveraging the measured melt pool thermal history and two newly created deep learning neural networks. A PyroNet, based on Convolutional Neural Networks, is developed to correlate in-process pyrometry images with layer-wise porosity; an IRNet, based on Long-term Recurrent Convolutional Networks, is developed to correlate sequential thermal images from an infrared camera with layer-wise porosity. Predictions from PyroNet and IRNet are fused at the decision level to obtain a more accurate prediction of layer-wise porosity. The model fidelity is validated with LBAM Ti-6Al-4V thin-wall structure. This is the first work that manages to fuse pyrometer data and infrared camera data for metal AM. The results of the case study show that our method can achieve 100% accuracy with relatively high efficiency, allowing the method to be applicable for in-situ porosity detection in LBAM.

A Physics-Informed Convolutional Neural Network with Custom Loss Functions for Porosity Prediction in Laser Metal Deposition

“…One challenge of integrating PIML into multimodal sensor data is matching modalities of different sensors. In one way, Guo et al appended finite element analysis (FEA)-simulated data to a convolutional neural network (CNN) model that incorporated thermal images from sensors [22]. In another way, Gawade et al matched the FEA-simulated melt pool data to thermal melt pool images by similar location and time [23].…”

“…Khanzadeh et al used morphological characteristics of the melt pool boundary to investigate the relationship between the melt pool characteristics and the defect occurrence in an as-built AM part [24]. Tian et al developed a deep learningbased data fusion method for in situ porosity detection in laser-based AM [22]. Yang et al developed a CNN model to investigate how the melt pool can be characterized in real time for feedback control of a laser powder bed fusion AM process [25].…”

“…CNNs can effectively reduce many parameters without compromising quality and are well suited to the given image classification task, where the images are high dimensional data. This baseline model uses the same PyroNet model as [22]. In particular, a VGG16 architecture [37] is chosen as the foundation of this model for its small 3 × 3 convolution cores, which increase the depth of the network and allow it to identify more complex patterns.…”

Application of nonlinear ultrasonic analysis for in situ monitoring of metal additive manufacturing

“…21 Recently, several methods have been designed for real-time porosity inspection during AM by online monitoring of melt-pool features (e.g., temperature, geometry, and radiation intensity). 22–26 Although these methods can examine porosity with a high accuracy, a large amount of iterative testing is required to establish a new relationship between the melt pool features and porosity once the target AM process or material is changed. In addition, the authors proposed a femtosecond laser-based transient thermoreflectance (TTR) measurement system for in situ porosity inspection.…”

In-Situ Quality Assessment and Control in Additive Manufacturing Processes Using Laser Surface Profilometer and Deep Learning Techniques

“…Although this approach showed great achievements in abnormality detection, the manual selection of features and thresholds need to show generalizability to more complex and non-linear problems. On the other hand, Qi et al (Tian et al, 2021) suggested DL based feature extraction for a metal AM machine. Their setup consisted of a pyrometer working alongside a thermal camera and a 3D CT scan machine.…”