Layer-Wise Modeling and Anomaly Detection for Laser-Based Additive Manufacturing

Abstract

Additive manufacturing (AM) is a novel fabrication technique capable of producing highly complex parts. Nevertheless, a major challenge is the quality assurance of the AM fabricated parts. While there are several ways of approaching this problem, how to develop informative process signatures to detect part anomalies for quality control is still an open question. The objective of this study is to build a new layer-wise process signature model to characterize the thermal-defect relationship. Based on melt pool images, we propose novel layer-wise key process signatures, which are calculated using multilinear principal component analysis (MPCA) and are directly correlated with the layer-wise quality of the part. The resultant layer-wise quality features can be used to predict the overall defect distribution of a fabricated layer during the build. The proposed model is validated through a case study based on a direct laser deposition experiment, where the layer-wise quality of the part is predicted on the fly. The accuracy of prediction is calculated using three measures (i.e., recall, precision, and F-score), showing reasonable success of the proposed methodology in predicting layer-wise quality. The proposed quality prediction methodology enables online process correction to eliminate anomalies and to ultimately improve the quality of the fabricated parts.

In Situ Nondestructive Fatigue‐Life Prediction of Additive Manufactured Parts by Establishing a Process–Defect–Property Relationship

“…The method used by Seifi et al does not centralize the ranges before the transformation. [54] Using their methodology causes the melt pool images to lose valuable information at the peak of melt pools. The reason behind this fact is the different thermal distribution of cylinder thermal images compared to the thin wall.…”

“…A convolutional neural network (CNN) is a type of deep neural network with the frequent use of analyzing visual imagery that is mainly used for image and video recognition, image b) the method used in the current study. [54] classification, etc. CNN uses convolution, a specialized kind of linear operation, instead of the general matrix multiplication in at least one of the layers of the network.…”

“…Seifi et al have provided the LENS machine and camera setup details. [54] This dataset is a collection of thermal images captured by a co-axial pyrometer camera, carrying the main information about the build process. The laser has used a unidirectional scan pattern with 15 tracks per layer and a finishing contour path.…”

“…Figure 3. The comparison of the two spherical methods: a) the method used by Seifi et al,b) the method used in the current study [54]. …”

Review on adaptive control of laser-directed energy deposition

“…Chen et al 85 investigated the mechanism of cracks in LDED and reduced the sensitivity of cracks by reducing heat input. One team at Mississippi State University [103][104][105][106] proposed the layer-wise processing method of multilinear principal component analysis, which can extract lowdimensional features of the molten pool to detect abnormalities in the process. The relation between molten pool morphology and microstructure anomalies was established by a machine learning framework, and the appearance of pores was predicted by a simulation model.…”