An official website of the United States government
PurposeThe purpose of this study is to develop a deep learning framework for additive manufacturing (AM), that can detect different defect types without being trained on specific defect data sets and can be applied for real-time process control. Design/methodology/approachThis study develops an explainable artificial intelligence (AI) framework, a zero-bias deep neural network (DNN) model for real-time defect detection during the AM process. In this method, the last dense layer of the DNN is replaced by two consecutive parts, a regular dense layer denoted (L1) for dimensional reduction, and a similarity matching layer (L2) for equal weight and non-biased cosine similarity matching. Grayscale images of 3D printed samples acquired during printing were used as the input to the zero-bias DNN. FindingsThis study demonstrates that the approach is capable of successfully detecting multiple types of defects such as cracks, stringing and warping with high accuracy without any prior training on defective data sets, with an accuracy of 99.5%. Practical implicationsOnce the model is set up, the computational time for anomaly detection is lower than the speed of image acquisition indicating the potential for real-time process control. It can also be used to minimize manual processing in AI-enabled AM. Originality/valueTo the best of the authors’ knowledge, this is the first study to use zero-bias DNN, an explainable AI approach for defect detection in AM.
more »
« less
Synthetic data augmentation to enhance manual and automated defect detection in microelectronics
Failure analysis and defect detection are crucial processes in industries, governments, and societies to mitigate the risks associated with defective microelectronics. The accurate identification of faulty parts is vital for preventing potential damages. However, traditional manual and automated defect detection approaches face challenges due to the scarcity of ground truth data from defective parts. This limitation hampers the effectiveness of subject matter experts and machine learning models in recognizing and classifying new instances of defects. To address this issue, we propose a synthetic data augmentation workflow that generates virtual defective parts, effectively overcoming the data scarcity problem and enabling the creation of large datasets at a low cost. Our approach enhances defect detection capabilities, empowering industries and governments to improve the quality and reliability of electronic devices.
more »
« less
- Award ID(s):
- 1916756
- PAR ID:
- 10542910
- Publisher / Repository:
- Elsevier
- Date Published:
- Journal Name:
- Microelectronics Reliability
- Volume:
- 150
- Issue:
- C
- ISSN:
- 0026-2714
- Page Range / eLocation ID:
- 115220
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Surface defect identification is a crucial task in many manufacturing systems, including automotive, aircraft, steel rolling, and precast concrete. Although image-based surface defect identification methods have been proposed, these methods usually have two limitations: images may lose partial information, such as depths of surface defects, and their precision is vulnerable to many factors, such as the inspection angle, light, color, noise, etc. Given that a three-dimensional (3D) point cloud can precisely represent the multidimensional structure of surface defects, we aim to detect and classify surface defects using a 3D point cloud. This has two major challenges: (i) the defects are often sparsely distributed over the surface, which makes their features prone to be hidden by the normal surface and (ii) different permutations and transformations of 3D point cloud may represent the same surface, so the proposed model needs to be permutation and transformation invariant. In this paper, a two-step surface defect identification approach is developed to investigate the defects’ patterns in 3D point cloud data. The proposed approach consists of an unsupervised method for defect detection and a multi-view deep learning model for defect classification, which can keep track of the features from both defective and non-defective regions. We prove that the proposed approach is invariant to different permutations and transformations. Two case studies are conducted for defect identification on the surfaces of synthetic aircraft fuselage and the real precast concrete specimen, respectively. The results show that our approach receives the best defect detection and classification accuracy compared with other benchmark methods.more » « less
-
Cracks and pores are two common defects in metallic additive manufacturing (AM) parts. In this paper, deep learning-based image analysis is performed for defect (cracks and pores) classification/detection based on SEM images of metallic AM parts. Three different levels of complexities, namely, defect classification, defect detection and defect image segmentation, are successfully achieved using a simple CNN model, the YOLOv4 model and the Detectron2 object detection library, respectively. The tuned CNN model can classify any single defect as either a crack or pore at almost 100% accuracy. The other two models can identify more than 90% of the cracks and pores in the testing images. In addition to the application of static image analysis, defect detection is also successfully applied on a video which mimics the AM process control images. The trained Detectron2 model can identify almost all the pores and cracks that exist in the original video. This study lays a foundation for future in situ process monitoring of the 3D printing process.more » « less
-
Vladislav Sergeevich Sorokin (Ed.)Phononic crystals can develop defects during manufacturing that alter the desired dynamic response and bandgap behavior. This frequency behavior change can enable successful defect inspection if the characteristic defect response is known. In this study, the behavior of a defective square unit cell comprising a freed and shortened leg is studied using a wave finite element method and an approximate continuous-lumped model to elucidate the defect induced qualitative dynamical features. These metrics are a computationally inexpensive alternative to modeling a defective unit cell within a large pristine array entirely in finite elements. The accuracy of these models is validated by comparing the result to a full finite element model. The impact of a shortened unit cell leg on the behaviors of an infinite array of defective cells and a finite array with a single defect are successfully predicted through dispersion curves and frequency response functions, respectively. These methods reveal defect-induced modes that split the local resonance bandgap of the pristine cell, as well as new anti-resonances resulting from the shortened leg. The study uses both approaches to evaluate the effect of defects in complex phononic crystal geometries and provides a comparative evaluation of the results of each model.more » « less
-
Accurate yield prediction in integrated circuit manufacturing enables accurate estimation of production cost and early detection of processing problems. It is known that defects tend to be clustered and a chip is likely to be defective if its neighbors are defective. This neighborhood effect is not well captured in traditional yield modeling approaches. We propose a new yield prediction model, called adjacency-clustering which addresses, for the first time, the neighborhood effect, and delivers prediction results that are significantly better than state-of-the-art methods. Adjacency-clustering (AC) model is a form of the Markov random field minimum energy model (MRF) that is primarily known in the context of image segmentation. AC model is a novel use of MRF for identifying defect patterns that enable diagnosis of failure causes in the manufacturing process. In this paper we utilize the defect patterns obtained by the AC model for yield prediction. We compare the performance of the AC model to that of leading yield prediction models including the Poisson, the negative binomial, the Poisson regression, and negative binomial regression models, on real data sets and on simulated data sets. The results demonstrate that the adjacency-clustering model captures the neighborhood effect and delivers superior prediction accuracy. Moreover, the concept and methodology of adjacency-clustering are not limited to integrated circuit manufacturing. Rather, it is applicable in any context where a neighborhood effect is present, such as disease risk mapping and energy consumption prediction. The e-companion is available at https://doi.org/10.1287/opre.2018.1741 .more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1016/j.microrel.2023.115220
