Flaw detection in non-destructive testing, especially for complex signals like ultrasonic data, has thus far relied heavily on the expertise and judgement of trained human inspectors. While automated systems have been used for a long time, these have mostly been limited to using simple decision automation, such as signal amplitude threshold. The recent advances in various machine learning algorithms have solved many similarly difficult classification problems, that have previously been considered intractable. For non-destructive testing, encouraging results have already been reported in the open literature, but the use of machine learning is still very limited in NDT applications in the field. Key issue hindering their use, is the limited availability of representative flawed data-sets to be used for training. In the present paper, we develop modern, deep convolutional network to detect flaws from phased-array ultrasonic data. We make extensive use of data augmentation to enhance the initially limited raw data and to aid learning. The data augmentation utilizes virtual flaws—a technique, that has successfully been used in training human inspectors and is soon to be used in nuclear inspection qualification. The results from the machine learning classifier are compared to human performance. We show, that using sophisticated data augmentation, modern deep learning networks can be trained to achieve human-level performance.

Additive manufacturing (AM) is a rapidly growing field of technology. In order to increase the variety of metal alloys applicable for AM, selective laser melting (SLM) of duplex stainless steel 2205 powder and the resulting microstructure, density, mechanical properties, and corrosion resistance were investigated. An optimal set of processing parameters for producing high density (>99.9%) material was established. Various post-processing heat treatments were applied on the as-built predominantly ferritic material to achieve the desired dual-phase microstructure. Effects of annealing at temperatures of 950 °C, 1000 °C, 1050 °C, and 1100 °C on microstructure, crystallographic texture, and phase balance were examined. As a result of annealing, 40–46 vol.% of austenite phase was formed. Annealing decreased the high yield and tensile strength values of the as-built material, but significantly increased the ductility. Annealing also decreased the residual stresses in the material. Mechanical properties of the SLM-processed and heat-treated materials outperformed those of conventionally produced alloy counterparts. Using a scanning strategy with 66° rotation between layers decreased the strength of the crystallographic texture. Electrochemical cyclic potentiodynamic polarization testing in 0.6 M NaCl solution at room temperature showed that the heat treatment improved the pitting corrosion resistance of the as-built SLM-processed material.

For estimating the probability of detection (POD) in non-destructive evaluation (NDE), there are two standard methods, the so-called â versus a approach and the hit/miss approach. The two approaches have different requirements for the quality and quantity of input data as well as for the underlying NDE method. There is considerable overlap between the methods, and they have different limitations, so it is of interest to study the differences arising from using each methodology. In particular, if the dataset is not ideal, the methodologies may exhibit different problems dealing with various limitations in the data. In this paper, a comparison between â versus a and hit/miss analysis was completed for two different data sets, a manual aerospace eddy-current inspection and a nuclear industry phased array ultrasonic weld inspection using a simplified online tool. It was found that the two standard methods (â vs. a and hit/miss) may give significantly different results, if the true hit/miss decision is based on inspector judgement and not automated signal threshold. The true inspector hit/miss performance shows significant variance that is not attributable to signal amplitude. Model-assisted POD was not able to model the inspector performance due to lack of representative amplitude threshold and difficulties in capturing true signal variance. The paper presents experience from practical cases and may be considered a European viewpoint.

Previous research (Li et al., Understanding the disharmony between dropout and batch normalization by variance shift. CoRR abs/1801.05134 (2018).http://arxiv.org/abs/1801.05134 arXiv:1801.05134) has shown the plausibility of using a modern deep convolutional neural network to detect flaws from phased-array ultrasonic data. This brings the repeatability and effectiveness of automated systems to complex ultrasonic signal evaluation, previously done exclusively by human inspectors. The major breakthrough was to use virtual flaws to generate ample flaw data for the teaching of the algorithm. This enabled the use of raw ultrasonic scan data for detection and to leverage some of the approaches used in machine learning for image recognition. Unlike traditional image recognition, training data for ultrasonic inspection is scarce. While virtual flaws allow us to broaden the data considerably, original flaws with proper flaw-size distribution are still required. This is of course the same for training human inspectors. The training of human inspectors is usually done with easily manufacturable flaws such as side-drilled holes and EDM notches. While the difference between these easily manufactured artificial flaws and real flaws is obvious, human inspectors still manage to train with them and perform well in real inspection scenarios. In the present work, we use a modern, deep convolutional neural network to detect flaws from phased-array ultrasonic data and compare the results achieved from different training data obtained from various artificial flaws. The model demonstrated good generalization capability toward flaw sizes larger than the original training data, and the effect of the minimum flaw size in the data set affects the $$a_{90/95}$$a90/95 value. This work also demonstrates how different artificial flaws, solidification cracks, EDM notch and simple simulated flaws generalize differently.

The scope of this research is to characterize and optimize the vibration-assisted ball burnishing of additively manufactured 18% Nickel Maraging steel for tooling applications. We evaluate the suitability of vibration-assisted ball burnishing as an alternative method to post-process additively manufactured tool steel. To do so, we assessed a single pass post-processing technique to enhance surface roughness, surface micro-hardness, and residual stress state. Results show that ultrasonic burnishing after age hardening functionalizes additively manufactured surfaces for tooling applications creating a beneficial compressive residual stress state on the surface. The surface micro-hardness (HV1) varied between 503 and 630 HV1, and the average surface roughness (Ra) varied between 1.31 and 0.14 µm, depending on process parameters with a maximum productivity rate of 41.66 cm²/min making it an alternative approach to functionalize additively manufactured tool components.

Dissimilar metal welds (DMWs) are a key design feature in nuclear power systems, typically involving ferritic low-alloy steels (LAS), stainless steels (SS), and nickel-base alloys. They are, however, a potential concern regarding the structural integrity of nuclear power systems. In particular, the LAS/nickel-base alloy weld metal interface is known to develop a local strength mismatch upon post-weld heat treatment (PWHT). Very limited data is available on the effect of thermal ageing on the DMW interface. The aim of this study was to investigate the effects of thermal ageing at 400 °C for up to 10,000 h on a narrow-gap DMW mock-up representative of the weld between the reactor pressure vessel nozzle and its safe-end after PWHT, with a special focus on the LAS SA 508/nickel-base Alloy 52 weld metal interface. No significant effect of thermal ageing on the appearing microstructure was observed in either LAS base material, LAS heat-affected zone or Alloy 52 weld metal. However, thermal ageing reduced the local strength mismatch at the LAS/nickel-base weld metal interface formed during PWHT. The reduction of the strength mismatch was detected using nanoindentation measurements and was concluded to be associated with a decrease in the carbon pile-up in the weld metal caused by PWHT. Based on the obtained results, thermal ageing promotes carbon diffusion from the weld metal side of the fusion boundary further away into the weld metal and thus slightly decreases the local strength mismatch.