BAND: BAgging noise detectors with application to semiconductor wafer denoising

The detection of process problems and parameter drift at an early stage is crucial to successful semiconductor manufacture. The defect patterns on the wafer can act as an important source of information for quality engineers allowing them to isolate production problems. Traditionally, defect recognition is performed by quality engineers using a scanning electron microscope. This manual approach is not only expensive and time consuming but also it leads to high misidentification levels. In this paper, an automatic approach consisting of a spatial filter, a classification module and an estimation module is proposed to validate both real and simulated data. Experimental results show that three types of typical defect patterns: (i) a linear scratch; (ii) a circular ring; and (iii) an elliptical zone can be successfully extracted and classified. A Gaussian EM algorithm is used to estimate the elliptic and linear patterns, and a spherical-shell algorithm is used to estimate ring patterns. Furthermore, both convex and nonconvex defect patterns can be simultaneously recognized via a hybrid clustering method. The proposed method has the potential to be applied to other industries.

Wafer Defect pattern recognition is an essential process in semiconductor manufacturing. This helps the manufacturers improve their fabrication process, reduce initial costs, and avoid further defects. However, mixed-type defects can be varied and complicated, which is difficult for traditional deep learning model to learn their features. To address this problem, we utilize dynamicity to allow the adaptive integration of information. Therefore, this paper proposed a mixed-type wafer defect pattern recognition framework based on multi-faceted dynamic convolution (MFD). This framework can effectively detect decisive features and integrate helpful information from defect patterns. Specifically, the proposed MFD firstly designs a 1-D auto-encoder to process the input, and extends it to a 3-D to obtain trans-pixel dynamic perception. In addition, we apply a channel-amalgamating mechanism to obtain the trans-channel dynamic perception and further synthesize the information across the channel. We appraised our model based on a dataset with 38 defect patterns generated from real industrial manufacturing. The results showed that the proposed framework excelled the vanilla CNN and state-of-the-art deep learning models. Further interpretable analysis based on visualization also indicated that this framework can adaptively procure the significant features from different defects.

Spectral Clustering (SC) is an effective clustering method for its excellent performance in partitioning non-linearly distributed data. On the other hand, Ensemble Clustering (EC), a different clustering technology, can promote cluster quality by ensembling the results of base clusterings. In this work, we concentrate on an EC framework that utilizes SC as the base method. Nevertheless, SC suffers from scalability due to its high computational complexity in constructing the Laplacian graph and computing the corresponding eigendecomposition. In the past decades, many efforts have been made to it. However, SC suffers from the scalability issue in processing extensive data, especially in web-scale scenarios. Additionally, EC requires multiple clustering results as the ensemble bases, which further aggravates resource consumption. To address this issue, LiteWSEC, a simple yet efficient Lightweight Framework for Web-scale Spectral Ensemble Clustering, is proposed to cluster web-scale data with limited resource requirements. It adopts the Web-scale Spectral Clustering (WSC) as the base method, which has minimal space overhead without computing overall embedding explicitly. LiteWSEC is highly flexible in the memory requirement, which is adaptive to the available resource. It can partition web-scale data (e.g., <inline-formula><tex-math notation="LaTeX">$n $</tex-math></inline-formula> = 8,000 k) in an resource-limited host (e.g., memory is restricted to 1 GB). Experiments on real-world, large-scale, and web-scale datasets demonstrate both the efficiency and effectiveness of LiteWSEC over state-of-the-art SC and EC methods.

Failing to release unneeded system resources such as I/O streams can result in resource leaks, which can lead to performance degradation and system crashes. Existing resource-leak detectors are usually based on predefined defect patterns to detect resource leaks in software. However, they typically report too many false positives and negatives, and also lack effective warning prioritization. Our empirical investigation shows that, their predefined defect patterns are not precise enough, and moreover, their used defect detection processes are not suitable enough for the defect patterns. In our approach, we introduce a novel Expressive Defect Pattern Specification Notation (EDPSN). With EDPSN, a resource-leak defect pattern can be defined more precisely by specifying conditional method calls and more expressively by including guiding information for the defect detection and warning prioritization process, such as the characteristics of its preferred defect detection process and the effective prioritization impact factors for its related warnings. Based on the EDPSN-based defect pattern, our approach tries to flexibly tune out a suitable defect detection and warning prioritization process. Through evaluations on three real-world projects (Eclipse-3.0.1, JBoss-3.0.6, and Weka-3.6.4), we show that our approach achieves high average precision (96%) and recall (74%), 26% and 49% higher than existing approaches, respectively.

To improve the performance of the adhesive layer defect detection (DD) planar array capacitor system, an adhesive layer DD method based on particle swarm optimization and fuzzy C-means (FCM) clustering algorithm is proposed. By classifying erroneous capacitance data, the experiment distinguishes the capacitance values at defect locations from those at defect-free locations. In response to the presence of artifacts in the image reconstruction results, the optimal threshold method is used to post-process the reconstructed images using the mixed regularization method. These experiments confirmed that FCM clustering reduced the mean absolute deviation (MAD) of data by about 6.1%, and the proposed clustering algorithm reduced the MAD of data by about 23.2%. Compared to the Tikhonov method, the mixed regularization method reduced the relative error of reconstructed images by about 13.1% and increased the correlation coefficient of reconstructed images by about 21.6%. Therefore, this research method had significant effects in reducing data fluctuations and errors, effectively improving the image reconstructing quality and the accuracy of DD. This study can effectively improve the accuracy of adhesive layer DD, help reduce production costs, and improve production efficiency.

Automated peanut defect detection using hyperspectral imaging and deep learning: A real-time approach for smart agriculture

Network Intrusion Detection Systems (NIDSs) are an increasingly important tool for the prevention and mitigation of cyber attacks. Over the past years, a lot of research efforts have aimed at leveraging the increasingly powerful models of Machine Learning (ML) for this purpose. A number of labelled synthetic datasets have been generated and made publicly available by researchers, and they have become the benchmarks via which new ML-based NIDS classifiers are being evaluated. Recently published results show excellent classification performance with these datasets, increasingly approaching 100 percent performance across key evaluation metrics such as Accuracy, F1 score, AUC, etc. Unfortunately, we have not yet seen these excellent academic research results translated into practical NIDS systems with such near-perfect performance. This motivated our research presented in this paper, where we analyse the statistical properties of the benign traffic in three of the more recent and relevant NIDS datasets, (CIC_IDS, UNSW_NB15, TON_IOT), by converting them into a common flow format. As a comparison, we consider two datasets obtained from real-world production networks, one from a university network and one from a medium size Internet Service Provider (ISP). Our results show that the two real-world datasets are quite similar among themselves in regards to most of the considered statistical features. Equally, the three synthetic datasets are also relatively similar within their group. However, and most importantly, our results show a distinct difference of most of the considered statistical features between the three synthetic datasets and the two real-world datasets. Since ML relies on the basic assumption of training and test datasets being sampled from the same distribution, this raises the question of how well the performance results of ML-classifiers trained on the considered synthetic datasets can translate and generalise to real-world networks. We believe this is an interesting and relevant question which provides motivation for further research in this space.

Capacitive touch panels (CTPs), as a medium of information interactions, have become essential parts in many consumer electronics. However, current methods such as image edge matching and frequency notch filter cannot suit the defect detection for the new-type complex CTP patterns, which have neither basic primitives nor periodicity. For solving the issues, we proposed a nonnegative matrix factorization (NMF)-based large-size image registration method, and combined it with image tolerance models to detect defects in such CTP patterns. The NMF-based image registration method can fast extract each CTP from a large image. And then, any three of registered images are selected as reference images, which are further processed by threshold processing and simple mathematical morphological operation to obtain tolerance models. Afterward, we can use the tolerance models to obtain a nondefective template. In the normal inspection stage, the defects in CTP patterns can be identified as long as comparing the tolerance models of the template and sensed images. The experimental results show that the proposed method can efficiently and accurately detect various types of defects in CTP patterns. Moreover, the detection results are robust under different illuminations. Therefore, this algorithm can be reliably applied in actual inspection of such new-type CTP patterns.

In this paper, we propose an accurate automatic defect detection method using the quadtree decomposition on scanning electron microscope (SEM) images. The proposed method consists of five steps including alignment between the inspection and reference images, edge relaxation, defect candidate detection, multichannel integration, and defect scoring. Since the periodicity of semiconductor SEM images is usually high, alignment can be failed due to a lot of local minima. To resolve this problem, we propose the enhancement method of non-periodic areas in SEM images. Then, we propose an edge relaxation technique to minimize the falsely remained edges in the aligned and subtracted image due to different focusing and exposure between two images. Then, the defect candidates are detected by analyzing the inhomogeneity based on quadtree decomposition, which is enhanced by our outlier-robust decomposition rule and adaptive decomposition thresholding to prevent over/under decomposition. Whereas the conventional defect detection system focuses only on the intensity of defects, our method considers both the inhomogeneity and intensity of defects. Field test results demonstrated that the proposed method provided more accurate defect detection with faster speed as compared to the conventional system. The proposed method can much reduce the automatic defect review rework compared with the conventional system.

Defect pattern classification of wafer bin maps (WBMs) assists in identifying the causes of semiconductor manufacturing process failures, thus contributing to the search for appropriate solutions. However, new previously unobserved defect patterns may be generated in real applications, which require classification methods with out-of-distribution (OOD) pattern detection capabilities. In this paper, we propose a novel method called a spectral-normalized neural radial basis function (SNRBF) network to classify defect patterns in WBMs while simultaneously detecting OOD defect patterns by quantifying the predictive uncertainty of a deep neural network. The SNRBF network adds spectral normalization to the weights in each layer of a deep neural network and replaces the output layer with a radial basis function kernel, facilitating high-quality uncertainty quantification and efficient computation. Additionally, we propose a novel regularization method based on singular values of the representation matrix to improve uncertainty quantification. Furthermore, we propose a new noise-filtering method capable of effectively eliminating random defects in WBMs by considering various defect pattern sizes. The proposed method is evaluated on a real WBM dataset and is demonstrated to outperform other competing methods in OOD detection while presenting competitive classification performance.

To increase the productivity, it is important to manage yield and reduce defects in the semiconductor industry. One of the efforts is to identify defect patterns and control the cause factors that affects the defects. Many engineers inspect the quality of each chip and check the defect pattern on the wafer bin maps. To get the accurate and consistent classification results regardless of the level for domain knowledge or experience of engineers, deep learning-based models have recently been studied. Since most previous studies aim to classify the single-type defect patterns, it is needed to consider the mixed-type defect patterns together. Also, they require a lot of labeled data to train the deep learning-based classification model. However, defects occur extremely rarely in actual manufacturing process. Therefore, the method securing the higher accuracy in a situation where enough labeled data are not given is needed. This paper proposes a deep convolutional generative adversarial network for wafer map synthesis (DCGAN-WS) which generates the mixed-type patterns by synthesizing the single-type pattern and adding the pixel-wise summation. To maintain the characteristics of the binary pixel of the wafer bin maps, a thresholding technique is added. MixedWM38 dataset is used for the experiments, and it was verified that the mixed-type patterns were synthesized well. It helps to construct more robust model for single-type pattern classification and to generate the mixed-type patterns that have not occurred before. In the future, it is expected that this model addresses the problem of the lack of labeled data for defect pattern classification models.

Constrained clustering has been well-studied for algorithms like K-means and hierarchical agglomerative clustering. However, how to encode constraints into spectral clustering remains a developing area. In this paper, we propose a flexible and generalized framework for constrained spectral clustering. In contrast to some previous efforts that implicitly encode Must-Link and Cannot-Link constraints by modifying the graph Laplacian or the resultant eigenspace, we present a more natural and principled formulation, which preserves the original graph Laplacian and explicitly encodes the constraints. Our method offers several practical advantages: it can encode the degree of belief (weight) in Must-Link and Cannot-Link constraints; it guarantees to lower-bound how well the given constraints are satisfied using a user-specified threshold; and it can be solved deterministically in polynomial time through generalized eigendecomposition. Furthermore, by inheriting the objective function from spectral clustering and explicitly encoding the constraints, much of the existing analysis of spectral clustering techniques is still valid. Consequently our work can be posed as a natural extension to unconstrained spectral clustering and be interpreted as finding the normalized min-cut of a labeled graph. We validate the effectiveness of our approach by empirical results on real-world data sets, with applications to constrained image segmentation and clustering benchmark data sets with both binary and degree-of-belief constraints.

Spectral clustering has become one of the most popular clustering methods due to its superior performance compared to the traditional clustering methods. However, the performance of spectral clustering would be limited by complex data, such as a huge number of samples and high dimensionality. To address this problem, some existing methods apply deep learning to learn the lower-dimensional representations, spectral clustering is then applied to the representations. Different from the existing methods that separate the two stages of feature representation learning and spectral clustering, in this paper, we propose Spectral Clustering Joint Deep Embedding (SCJDE), a method that simultaneously learns the feature representations and the spectral embedding of spectral clustering via a deep autoencoder. Moreover, a sparsity constraint is imposed to generate better spectral embedding for spectral clustering. Finally, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$k$</tex> -means is performed on the spectral embedding to obtain clustering result. The proposed method can learn a good spectral embedding for spectral clustering by deep learning to obtain better clustering results. The experimental results on both synthetic and real-world datasets demonstrate the effectiveness of the proposed method.

Research on the fairness of spectral clustering has gradually increased attention. Normally, existing methods of fair spectral clustering add a fairness constraint to the original objective function so that fairness is guaranteed. However, similar to the solver of traditional spectral clustering, that of fairness spectral clustering has to relax a discrete value condition into an arbitrary one, which leads to the deterioration of both fairness and clustering quality. Moreover, the eigen-problem is inevitable in the solver, which takes O(n3) time complexity and is not available for large-scale data. In this paper, we propose a fair spectral clustering algorithm by employing the coordinate descent method to find the solution. As the relaxation of the discreteness condition is discarded, the fairness is improved. Furthermore, we refine the process of coordinate descent by avoiding redundant calculations, and as a result, the time complexity is reduced from O(n3) to O(n2). Additionally, the importance of clustering quality and fairness is symmetric; hence, we achieve a trade-off between them by adjusting the parameters. The experimental findings, obtained from both real-world and synthetic datasets, clearly illustrate that our proposal delivers superior fairness and clustering quality with the best BAL compared to other fair clustering methods. In addition, our method is more efficient than existing fair spectral clustering algorithms.

To explore the size, morphology, and distribution patterns of internal pore defects in WE43 magnesium alloy formed by laser powder bed fusion (LPBF), as well as their impact on its mechanical properties, computer tomography (CT), metallographic microscopy, and scanning electron microscopy were used to observe the material's microstructure and the morphology of tensile test fractures. The study revealed that a large number of randomly distributed non-circular pore defects exist internally in the LPBF-formed WE43 magnesium alloy, with a defect volume fraction of 0.16%. Approximately 80% of the defects had equivalent diameters concentrated in the range of 10∼40 μm, and 56.2% of the defects had sphericity values between 0.65∼0.7 μm, with the maximum defect equivalent diameter being 122 μm. There were a few spherical pores around 20 μm in diameter in the specimens, and unfused powder particles were found in pore defects near the edges of the parts. Under the test conditions, the fusion pool structure of LPBF-formed WE43 magnesium alloy resembled a semi-elliptical shape with a height of around 66 μm, capable of fusion three layers of powder material in a single pass. Columnar grains formed at the edge of individual fusion pools, while the central area exhibited equiaxed grains. The "scale-like pattern" formed by overlapping fusion pool structures resulted in the microstructure of LPBF-formed WE43 magnesium alloy mainly consisting of fine equiaxed grains with a size of 2.5 μm and columnar grains distributed in a band-like manner.