Papers by Keyword: Defect Detection

Abstract: The increasing complexity of X-ray test data and the demand for precise, large-scale evaluations have exposed the limitations of manual flaw detection methods in Non-Destructive Testing (NDT). Traditional manual approaches, while widely used, are labor-intensive and prone to human error, often leading to inconsistencies and inefficiencies. An Artificial Intelligent driven (AI-driven) system has been developed to automate the analysis and evaluation of X-ray test data, improving both accuracy and efficiency. The system employs advanced image segmentation and classification algorithms, aligning with ISO standards to detect defects in welded structures. By reducing reliance on manual interpretation, the system enhances the reliability and speed of the evaluation process, while also allowing for expert oversight through manual corrections when necessary. Integration with a Laboratory Information Management System (LIMS) ensures streamlined data handling and traceability, minimizing human error in data recording. As the AI model processes more data, it continuously improves, adapting to evolving defect patterns and maintaining high performance. This paper details the system’s AI architecture, the methodology employed for X-ray image analysis, and the performance results from industrial applications, demonstrating how this technology addresses key challenges in NDT processes.

Abstract: We have developed methodologies for the detection of microdefects on painted surfaces and the evaluation of machined surface quality through the use of the patterned area illumination. The objective of this study is to optimize illumination for surface inspection of products by reproducing the illumination situation using the patterned area illumination. In this study, a ray tracing method, which is a computer graphics technique, was employed to reproduce a scenario in which a product surface is illuminated by the aforementioned patterned area illumination. In the conventional ray tracing method, the patterned area illumination is not taken into account. Accordingly, the surface to be inspected was postulated to be a mirror surface, and the illumination pattern was projected onto it in order to reproduce the reflection of the illumination pattern on the inspection surface. The developed simulation method enables several key optimizations. Firstly, it allows for the optimization of the illumination device by reproducing the inspection surface with a curved surface. Secondly, it enables the data augmentation of teacher data for machine learning for a versatile defect detector. Thirdly, it allows for the optimization of the circular pattern used to estimate the shape of micro defects. Finally, it enables the reproduction of pattern projection onto a machined surface for the evaluation of machining quality of the machined surface.

Abstract: This paper explores advanced methods and techniques for defect detection, focusing on their effectiveness, challenges, and implications for industrial applications. We explore the combination of CNNs with deflectometry and dark-field polarization imaging for surface defect detection in refrigerator manufacturing and optical components inspection, respectively. We highlight the importance of automated inspection systems in detecting surface defects and discuss the challenges associated with real-time defect detection and limited datasets. This study contributes to advancing defect detection methodologies and provides valuable insights for industrial quality control processes.

Abstract: Typically, research and development (R&D) results of epitaxial layer growth show superior properties of the grown layers compared to high volume results. Layer uniformities are excellent and achieved defect densities are low compared to typical results. In particular, the conversion of basal plane dislocations (BPD) from the silicon carbide (SiC) substrate is in focus to reduce bipolar degradation of p-n-junctions. It is a great challenge to maintain those excellent results in high-volume manufacturing considering all the factors that impact the properties of the epilayer. Thus, quality of the layers, high throughput and low cost have to be assessed to find a compromise between these key factors. In this paper we present results on the growth of epitaxial layers on 150 mm and 200 mm 4° off-oriented 4H-SiC substrates using warm-wall multi-wafer chemical vapor deposition (CVD) systems. Single wafer data of the key epitaxial layer parameters, thickness, doping and defect densities, are compared to batch and lot results, as well as to statistical data of several hundreds of wafers produced. Improvements in wafer-to-wafer (w-t-w) doping uniformity could be achieved for instance by implementation of an on-wafer temperature measurement. Substrate impact on defect levels is shown comparing X-ray topography (XRT) results of bare substrate wafers and defect analysis of epilayers on sister wafers from the same crystal. Statistical defect data and resulting predicted yield loss also show a dependence on substrate suppliers. For the first time we show w-t-w and run-to-run (r-t-r) results of doping and thickness measurements on 200 mm substrates. Also, defect results of epilayers on 200 mm wafers are compared to results on 150 mm.

Abstract: The paper proposed a deep convolutional neural network together with image processing techniques to detect assembly defects of vehicle components in assembly lines. Traditional detection method such as automatic optical inspection is strongly affected by environmental variation coming from the changes of light source, transfer belt, and component type, therefore, complicated thresholds should be adjusted case by case. The proposed method tries to avoid these problems which is fast and straight forward with satisfactory detection accuracy compared to traditional method.

Abstract: Nowadays new applications based on the 3D printing technique demand increasingly strict product quality requirements. The in-situ monitoring of variables associated with the manufacturing process through the application of different techniques could help to evaluate the process and ultimately to ensure product quality. In this regard, the acquisition and evaluation of variables and indexes derived from thermographic analysis during the process are key for an early defect detection and can contribute to quality estimation. In this work, a new methodology is proposed for the monitoring and analysis of the additive manufacturing process based on the processing of thermographic images from an LWIR (Long Wave Infrared) camera. The methodology and the suitability of the variables and indexes extracted during the monitoring of the manufacturing process are discussed for the case of a 3D fused filament fabrication of polymers.

Abstract: Welding processes are commonly used in the industry for the manufacture of large parts or due to their complex geometry that does not allow the part to be manufactured as a whole. Nevertheless, the union can show worst mechanical properties than the rest of the piece, affecting negatively its service behavior, so it is necessary evaluate weald seam to ensure the correct process application. Electrical welding operations are commonly used due to the reduced equipment size or their possibilities application in numerous metallic materials. Notwithstanding, different variables have to be taken into account during the metal deposition, as intensity or speed deposition, among others. Weald seam geometrical evaluation is usually utilized to validate the union surface conditions. Despite this, surface irregularities caused during the process make its difficult to measure correctly with conventional equipment. However, laser profilometry is a non-contact technique that can be used to generate 3D profile of weald seam, facilitating its measurement with high accuracy. Therefore, in this work an initial analysis of the influence of material deposition speed and arc welding intensity on the weald seam geometry will be carried out using a laser profilometry equipment. In addition, to ensure a correct information acquisition, the laser profilometer requires a constant speed movement in relation with the weald seam analyzed, so new equipment has been manufactured, using additive manufacturing techniques, to support the profilometer throughout the information acquisition process.

Abstract: The quality of the silicon carbide (SiC) epitaxial layer, i.e., layer homogeneities and extended defect densities, is of highest importance for high power 4H-SiC trench metal-oxide-semiconductor field effect transistors (Trench-MOSFET) devices. Especially, yield for devices with a large chip area is severely impacted by extended defects. Previously, devices had to be fully manufactured to effectively gauge the impact of a reduction in extended defect densities in the epitaxial layers on device yield. The production of devices such as Trench-MOSFETs is an extensive procedure. Therefore, a correlation between extended defects in the epitaxial layer and electrical device failure would allow to reliably estimate the impact of process changes during epitaxial layer deposition on electrical device yield.For this reason, n-type epitaxial layers were grown on around 1,000 commercially available 150 mm 4H-SiC Si-face substrates, which received a chemical wet cleaning prior to the epitaxy deposition. Substrates with lowest micro-pipe density from two different suppliers were used. The wafers were characterized with the corresponding device layout for defects utilizing surface microscopy as well as ultraviolet photoluminescence techniques. Subsequently, these wafers were used to produce more than 500,000 Trench-MOSFET devices. All devices have been tested on wafer level for their initial electrical integrity.With these methods a precise correlation between extended defects in the epitaxial layer and electrical failures on wafer level could be found. The influence of different substrates on the defect-based yield prediction regarding the electrical yield on wafer level is discussed. Additionally, a calculated kill-ratio is presented and the severity of defect classes on initial device failure, e.g., stacking faults, and their key failures modes are discussed.

Abstract: Nowadays, the industry requires automatic production for high-speed manufacturing. However, the products must also maintain high quality and reliability. An efficient inspection technique should be conducted for the improvement in the manufacturing quality. In order to achieve high inspection rate, optical inspection based on machine vision often raises the threshold of the judgment and this will worsen false detection. In this study, we propose a high-accuracy optical inspection system based on deep learning technology. Various defects in screw head are precisely detected and analyzed, which include surface damage, unprocessed, and stripped surfaces. An industrial camera and microscope system are employed to collect the raw images of metal screws with different defect types. The raw images of 3200 are utilized to train the designed convolutional neural networks. The experimental results indicate that the proposed system reaches a detection accuracy of 92.8% and the average detection speed is 0.03 second per image. In comparison with conventional machine vision methods, the proposed measurement system is more suitable for the inspection of industrial production line.

Abstract: Additive Manufacturing (AM) using Powder-Bed Fusion Laser-Beam (PBF-LB) has great potential; however, it has challenges due to its sensitivity to the process parameters [1]. The availability of big data generated in AM facilitates the employment of Machine Learning (ML) tools to understand the process and have a predictive control over the production. An intelligent system like this can thus reduce material wastage and energy cost while increasing a plant’s product quality and throughput. Time-series summary statistics (like mean and variance) can discard valuable discriminatory signatures embedded in raw sensor data. Therefore, special ML time-series classification (TSC) tools that can extract and utilise these signatures from the raw data are much more effective for a task like porosity prediction [1]. However, the data employed in [1] pertains to products with artificially designed pores or gaps. This study focuses on naturally occurring pores, rarer, and evaluates k-Nearest Neighbour (k-NN) with Dynamic Time Warping (DTW) over real-world manufacturing data to classify the porosity of individual raster scans. We believe that natural pores have more diverse signatures than artificial pores, as each pore varies in characteristics (like size and morphology).