Bayesian Network Machine Learning Approach to Atmospheric Corrosion Modelling

Atmospheric corrosion is strongly influenced by the persistence of surface moisture, commonly quantified through the time of wetness (TOW). Traditional TOW models, such as those defined by ISO 9223, rely on fixed humidity and temperature thresholds and often fail to capture the complexity of real-world environmental conditions. This study introduces a novel TOW modelling approach based on the concept of damage accumulation supported by machine learning. Using environmental data from ten monitoring stations across New Zealand, 61 features were developed to simulate conditions conducive to electrolyte formation. Feature importance analysis identified three key predictors: two condensation-related humidity factors and cumulative solar radiation. A simplified corrosion model based on these three predictors achieved strong predictive performance (R² = 0.909). Bayesian linear regression confirmed the statistical relevance and directional influence of the selected factors. Notably, the Condensation Factor and the Dilution Factor demonstrated opposing effects on corrosion rate. The Condensation Factor, linked to early-stage condensation and the formation of concentrated electrolytes in droplets, showed a positive correlation with corrosion. In contrast, The Dilution Factor, associated with intense wetting events, exhibited a negative correlation, possibly due to the dilution of surface contaminants. The created model outperforms current ISO’s TOW model to predict atmospheric corrosion rate. This framework highlights the role of moisture in atmospheric corrosion and provides a foundation for future integration with corrodent-specific models. The proposed methodology offers a scalable and interpretable tool for improving corrosion rate predictions in diverse environmental settings.

In colder climates, factors like substantial snowfall, frequent freeze-thaw cycles, the use of de-icing salts, and increasing global temperatures add complexities to the established paradigms governing atmospheric corrosion. To elucidate the mechanisms of atmospheric corrosion under colder temperatures, four distinct test sites were established in the state of Alaska, USA. Each site was equipped with a modern, multi-angle exposure test rack coupled with an auxiliary weather station. Within these setups, five commonly used metal alloys were exposed at 0-, 30-, and 45 degrees from the horizontal over 6- and 12-month periods. Weather correlation analysis, mass loss analysis, SEM/EDXA, and optical profilometry studies were conducted following the exposure of the metal specimens to determine corrosion rates (CR) and damage. The measured CR, exposure angle, and accompanying meteorological data—specifically, temperature, relative humidity, time of wetness, precipitation, chlorides, and sulfates—were benchmarked against ISO standards to classify and categorize the corrosivity of the four test sites in Alaska. The PSCA test site (Kodiak, AK) was the most aggressive environment regarding all weather parameters. UAF (Fairbanks, AK) was the least aggressive, and test sites PAA (Port of Anchorage, AK) and UAA (University of Anchorage, AK) showed mild to moderate aggressiveness. Samples exposed at 0° consistently showed higher corrosion rates than those exposed at 30° and 45° across all instances. At the more aggressive PSCA site, a distinct correlation was observed between the exposure angle and CR with samples exposed at 0° showing the highest CR, followed by 30°, then 45°. Optical profilometry studies revealed that the comparison of surface roughness between the exposed samples followed an extremely similar pattern to that of CR. Corrosivity categorization based on CR from experimental mass loss data varied from the estimation of CR through existing dose-response functions and weather parameters. Newer standardized models involving more weather parameters are required to better estimate and predict corrosivity in colder climates. Alaska's growing role as a critical military, cargo, shipping, and refueling hub enhances its importance as an ideal natural laboratory for studying atmospheric corrosion in cold regions, as demonstrated by the developed test sites discussed in this paper.

Study question Can an image recognition model, utilising automated deep learning, be trained to diagnose and classify the severity of adenomyosis on ultrasound? Summary answer The automated deep learning model developed in this study can accurately diagnose and classify the severity of adenomyosis from ultrasound images. What is known already Adenomyosis is prevalent across all stages of life, starting from adolescence to perimenopause, however it remains underdiagnosed. The diagnosis and classification of adenomyosis are essential for correlating with symptom severity and counselling about disease prognosis, management options and long-term impact. Diagnosing and classifying the severity is a time-consuming and complex process that depends on the operator’s experience. Machine learning (ML) has an enormous potential to assist healthcare professionals with image classification tasks. It can directly process and automatically learn mid to high-level abstract features acquired from images using a deep architecture model without requiring the manual definition of features. Study design, size, duration This observational cohort study utilised ultrasound images from four assisted conception units in the United Kingdom from February 2023 to February 2024. Labelled images were used as input for training and testing of the model. The complete dataset was manually split into 9:1 ratio; 90% of all images were used as training dataset (80% for training and 10% for validation) and 10% as test dataset. Ethics approval was obtained from the Institutional Review Board (IRB). Participants/materials, setting, methods An automated ML tool (Vertex AI Vision) was trained and tested to diagnose and classify adenomyosis severity. Two investigators, experienced in gynaecological ultrasound scanning, labelled two and three-dimensional images into four categories: normal uterus, mild, moderate, and severe adenomyotic uterus, as per standardised ultrasound criteria. The labelled images were formatted and uploaded on Google Cloud to train the neural network based on pattern recognition. The model’s performance was determined for an internal test dataset. Main results and the role of chance A total of 3,500 ultrasound images were obtained from 3,194 patients, and the internal test dataset consisted of 350 ultrasound images from 292 patients. An overall and per-category score threshold of 0.5 was adopted to provide a balance between reasonable precision and recall during the evaluation. The investigator’s categorization was used as the reference standard to evaluate the performance of the automated machine-learning model. The model’s overall accuracy for diagnosing and classifying adenomyosis, as determined by the area under the precision-recall curve, was 80.24% across all score thresholds. The overall precision was 83.00%, and recall was 65.75% at a confidence threshold of 0.5. The precision and recall were 91% and 78% for normal uterus, 67% and 53% for mild adenomyosis, 86% and 68% for moderate adenomyosis, and 88% and 64% for severe adenomyosis. The algorithm achieved a high accuracy for the internal dataset. The time taken by a trained healthcare professional to label and process 3,500 images into 4 categories was 70 hours. The time taken by an untrained automated ML tool to get trained, process, and label the images into 4 categories was 4 hours, saving 66 hours per healthcare professional. Limitations, reasons for caution Due to the inherent nature of non-invasive investigations for the population attending the assisted conception unit, histological confirmation of adenomyosis was not possible. In addition, the model’s performance has been tested using an internal test dataset and needs further validation using an external test dataset from a different population. Wider implications of the findings This novel study leverages the power of artificial intelligence to develop an automated machine learning model for diagnosing and classifying adenomyosis from ultrasound scans. This tool can potentially enhance the diagnostic capacity and efficiency of healthcare professionals and empower women by providing better care through early and uniform diagnosis. Trial registration number No

Corrosion under sodium chloride droplets on copper is being studied via electrochemical modeling and experimentally characterized under rigorously controlled laboratory conditions. The objective of this work is to understand the governance of thin alkaline films that emanate from saline droplets upon corrosion initiation on corrosion rates and damage distributions. The radial spreading of thin alkaline films has been reported for both macroscopic and microscopic saline droplets on a variety of alloys that exhibit active corrosion under humid conditions [1-6]. This process is termed secondary spreading. It is resultant from oxygen reduction at cathode sites near the drop edge which produce concentrated hydroxide solution (e.g., NaOH) that subsequently spreads as a film over the metal surface [1,2,7,8]. These films are characteristically comprised of fields of microdroplets on the order of 100 μm diameter or less and can spread several times the original droplet radius from which they emanated. A number of investigators have characterized secondary spreading, demonstrating the potential for these films to serve as sites for oxygen reduction [7]. Unknown is the extent to which these secondary spreading films provide cathodic current to the corroding system and, in turn, their influence on corrosion rates and damage distributions. In this work, we explore the degree to which secondary spreading films influence corrosion that occurs under NaCl droplets on copper for a range of droplet diameters (50 µm to 1 mm). Our approach is to (1) develop a continuum-based model of the system (with non-continuum informed model extensions) to explore the influence of the film under various conditions and (2) define the physicochemical properties of the droplet-film system and experimentally realize the influence of the film on corrosion behavior to inform the model. Specifically, a transient, multiphysics finite element model is being constructed to predict current and potential distributions during the initial stage of corrosion over a range of droplet-film geometries and electrochemical parameters. Initial model assumptions include Nernst-Planck transport kinetics in non-dilute solution and the absence of heterogeneous reactions in the bulk electrolyte; both assumptions will be improved upon in subsequent iterations. Additionally, Butler-Volmer electrochemical kinetics is assumed, with oxygen reduction as the only cathodic reaction and metal oxidation as the only anodic reaction. The behavior of the droplet-film system is being experimentally studied via electrochemical capillary techniques and the physicochemical evolution of the droplet and film characterized to inform and benchmark the model. Key to the experimental efforts is the ability to rigorously control electrolyte geometry. To do so, a novel testing platform has been constructed that will allow production of varying size microdroplets and allow electrochemical probing within them. Experiments are currently under way to characterize the effect of restricted size secondary spreading regions on corrosion kinetics and damage distributions using this system. Together, this integrated computational and experimental approach seeks to deliver new mechanistic insight into atmospheric aerosol-driven corrosion.

Large steel space structures, when exposed to a harsh corrosive environment, are inevitably subjected to atmospheric corrosion and stress corrosion cracking. This paper proposes a framework for assessing the corrosion damage of large steel space structures subjected to both stress corrosion cracking and atmospheric corrosion. The empirical model for estimating atmospheric corrosion based on measured information is briefly introduced. The proposed framework is applied to a real large steel space structure built in the southern coastal area in China to assess its corrosion damage and investigate the effects of atmospheric corrosion on stress corrosion cracking. Based on the results, the conceptual design of the corrosion monitoring system of large steel space structures is finally conducted as the first step for a real corrosion monitoring system.

Background Industrial content and infrastructure are in constant danger from atmospheric corrosion, which affects economies globally. However, there is a lack of a consistent set of comprehensive data that completely surrounds the range of this problem in diverse climate and locations. The purpose of the research is to evaluate the factors that contribute to atmospheric corrosion and its diverse effects on materials in various environments. Methods By creating a comprehensive dataset by collecting and standardizing corrosion data from diverse environments and geographic regions and initially analyzing the data, it helped indicate the main parameters affecting corrosion. This guided the selection of future features for further modeling. Several machine learning algorithms were tested, such as linear regression, decisions tree, neural network, and, most especially, attempting methods, for their corrosion rate prediction capabilities. These models were assessed based on their prediction’s accuracy, and computational efficiency, with special attention to refining their performance through detailed feature engineering and hyperparameter adjustment. Results Upon evaluating the performance of conventional predictive models, the research indicated that the machine learning approaches, especially with random forests methods of dress, were excellent in predicting corrosion rates, significantly improved upon these capabilities. By analyzing various machine learning approaches, it became clear that it was important to enhance their accuracy by selecting the best features and customizing them. Conclusions This work represents a significant advancement in the predictive modeling of atmospheric corrosion. It highlights the invaluable role of machine learning in this field. By integrating varied data sets and applying sophisticated machine learning techniques, it has established a foundation for ongoing research and the practical application of corrosion management strategies. The exceptional performance of ensemble methods, like random forests, signals their potential to improve prediction capabilities and guide more effective corrosion prevention measures.

The paper presents the results ofexperimental studies of the influence of corrosiondamage on bond of reinforcement in concrete. Experimental studies included tests for bond to concrete of reinforcement bars without corrosion damage and with corrosion damage in the form of rustcaused by atmospheric corrosion. Samples of reinforcement with corrosion damage were taken fromreinforcement stored in the open air for 10 monthsfrom January to October, and samples of reinforcement without corrosion damage - from fittingsstored in the production shop.Tests of reinforcement samples for bond toconcrete were performed by pulling reinforcementfrom concrete prisms, the length of which was 15d(d - diameter of the bar) with measurement in theprocess of testing the shear of the free end of thebar and subsequent plotting of tensile stress in thereinforcement offset of the free end of the bar.According to the results of tests of stable impact of the surface of the reinforcement - withoutcorrosion damage and with damage caused byatmospheric corrosion, not detected. The exceptionis the fittings of the smallest of diameters -Ø12A500C, for which the negative impact of corrosion damage in the range of 12… 16%. can beexplained by the fact that, other things being equal,the same amount of corrosion damage to a greaterextent affects the bond of valves of smaller diameters.

Copper and its alloys are highly valued for their moldability, heat/electricity conduction, and corrosion resistance, making them widely used in various industrial applications. One potential application of copper is as the corrosion barrier for used nuclear fuel containers (UFCs) within a deep geological repository (DGR) in Canada, Sweden and Finland. The UFC's copper surface will be exposed to different conditions in the evolving DGR environment, which will transition from warm, dry, and oxic to cool, wet, and anoxic conditions. Deliquescence is a known process in forming the initial corrosive solutions/droplets that may occur as a result of adsorption of moisture by salt particles at the UFC surface during the unsaturated phase in a DGR. The duration over which the surface will remain wet under oxic condition will impact the extent of copper corrosion. To investigate this phenomenon, the quartz crystal microbalance (QCM) technique was used. The QCM consists of a quartz plate, covered by gold electrodes on both sides onto which a layer of copper is deposited using e-beam deposition. In this technique, the change in the resonance frequency of a vibrating quartz crystal is linearly proportional to the change in mass on the crystal's surface (Δm) under certain conditions with a sensitivity in the nanogram range. In this study this effect was used to monitor the adsorption and evaporation of water, the salt depletion, and to quantify the extent of atmospheric corrosion on copper surfaces.To systematically study the atmospheric corrosion of copper surfaces, controlled amounts of different salts (NaCl, CaCl2, and MgCl2) were loaded onto copper-coated QCM crystals using an inkjet printer, and then corrosion tests were performed at different relative humidity (RH) values depending on the deliquescence relative humidity (DRH) of each salt at 75 °C.Results revealed that for NaCl, the surface remained continuously wet above its DRH (75% RH). However, for CaCl2, the time of wetness was relatively short post-deliquescence. Corrosion continued beyond this wet period, eventually depleting all the salt and forming non-deliquescent products. Higher RH levels prolonged the wetness period and accelerated the salt depletion, leading to a stabilized corrosion rate comparable to that of bare copper under similar conditions. Similarly for MgCl2, the time of wetness was also limited, and higher RH extended the wetness duration. Interestingly, corrosion rates differed from those observed with CaCl2. Surface analysis techniques, including focused ion beam (FIB) cutting and energy dispersive X-ray spectroscopy (EDS), were used to examine the composition and structure of corrosion films for CaCl2, providing insights into the underlying corrosion mechanisms. Further analyses are ongoing for both CaCl2 and MgCl2 to better understand their distinct corrosion behaviours.

Corrosion is widely known to be a major cause of the failures in process facilities. Prediction of corrosion damage is therefore essential for industries to manage the availability of their assets. This research aims to investigate the application of supervised machine learning methods for the classification of pitting corrosion damage. Several machine learning classifiers, namely ensemble methods, support vector machine (SVM), K‐nearest neighbours, and the decision tree are used to classify the extent of pitting corrosion damage in corroded steel samples. To simulate the corrosion of the steel samples, a series of laboratory experiments were conducted. After processing the results using appropriate statistical methods, the corrosion data was used to train the machine learning models. The trained models can predict the class of corrosion damage with acceptable accuracy using the material and environmental specifications of the samples. Additionally, a discussion on the selection of machine learning techniques which classify corrosion damage using a risk‐based approach is provided. With their optimal accuracy and lower risk of misclassification, the SVM and AdaBoost models perform better than the other studied models.

We present a physically based Bayesian network model for inference and prediction of flood duration that allows for a deeper understanding of the nexus of antecedent flow regime, atmospheric blocking, and moisture transport/release mechanisms. Distinct scaling factors at the land surface and regional atmospheric levels are unraveled using this Bayesian network model. Land surface scaling explains the variability in flood duration as a function of cumulative exceedance index, a new measure that represents the evolution of the flood in the basin. Dynamic atmospheric scaling explains the cumulative exceedance index using the interaction between atmospheric blocking system and the synergistic model of wind divergence and atmospheric water vapor. Our findings underline that the synergy between a large persistent low-pressure blocking system and a higher rate of divergent wind often triggers a long-duration flood, even in the presence of moderate moisture supply in the atmosphere. This condition in turn causes an extremely long-duration flood if the basin-wide cumulative flow prior to the flood event was already high. Thus, this new land-atmospheric interaction framework integrates regional flood duration scaling and dynamic atmospheric scaling to enable the coupling of ‘horizontal’ (for example, streamflow accumulation inside the basin) and ‘vertical’ flow of information (for example, interrelated land and ocean-atmosphere interactions), providing an improved understanding of the critical forcing of regional hydroclimatic systems. This Bayesian model approach is applied to the Missouri River Basin, which has the largest system of reservoirs in the United States. Our predictive model can aid in decision support systems for the protection of national infrastructure against long-duration flood events.

Improving diagnostic recognition of primary hyperparathyroidism with machine learning

The atmospheric corrosion of metals and alloys is a very complex process, which involves a wide range of chemical, electrochemical, and physical processes occurring in gas, liquid, and solid phases, as well as at the interfaces between them. Essentially all metals and alloys undergo corrosion in indoor or outdoor atmospheric conditions due to the presence of gaseous corrosion stimulators. However, the rate of the atmospheric corrosion might vary significantly depending on the nature of the metals or alloys, the concentration of corrosive species, as well as the exposure conditions. In some applications such as electrical contacts in microelectronics, military equipment or nuclear waste containers, even ultra-slow oxidation/corrosion may result in catastrophic consequences. In order to study such slow oxidation/corrosion processes, techniques involving ultra-high vacuum (UHV) are commonly used. Despite a very high surface sensitivity of these techniques they suffer from the drawback that the experiments are not performed under normal ambient conditions, but rather under vacuum. This has the consequence that the water adlayer which always exists on metal surfaces in atmospheric environments is removed under UHV conditions. Furthermore, in the case of ex-situ UHV studies, the structure of the oxidation/corrosion products might undergo vacuum induced changes. Hence, in-situ and highly surface sensitive techniques capable of working under atmospheric pressures are desired to monitor very slow oxidation and corrosion processes on metals and alloys at a molecular level. In the present study, several analytical techniques have been used to obtain a picture as complete as possible of an atmospheric corrosion process, focusing on both the formation of corrosion products and the evolution of the protective layer. Copper surfaces covered with protective organic monolayers have been exposed to various atmospheric environments with the aim of simulating typical indoor atmospheric corrosion processes. Infrared reflection/absorption spectroscopy (IRAS) was used to determine the nature of the corrosion products, quartz crystal microbalance (QCM) to measure the mass of them, and nano plasmonic sensing (NPS) to determine the corrosion rate as well as to evaluate how homogeneous the formation of the corrosion products was. In addition, an inherently surface sensitive technique, vibrational sum frequency spectroscopy (VSFS) was used to study induced changes in the molecular structure of the protective monolayers. VSFS studies of octadecanethiol (ODT) covered copper surfaces exposed to dry air allowed the detection of the formation of thin copper oxide layers and the observation of accompanying changes in the molecular structure of the hydrocarbon chains of ODT. VSFS was combined with QCM to quantify the amount of copper oxide formed on the surface. Since oxidation and corrosion processes often occur heterogeneously, the VSFS-QCM setup was integrated with NPS, which enabled us to distinguish homogeneous and heterogeneous corrosion under in-situ conditions with a time resolution on the order of seconds. This unique combination of three in-situ techniques facilitates the study of different corrosion processes where interfacial changes in electronic and optical properties as well as changes in mass occur. The sensitivity of the techniques under current exposure conditions corresponds to 5% of an oxide monolayer for VSFS, 2.5% for QCM, and 1.2% for NPS. As an extension of the above mentioned studies a combination of VSFS and IRAS was used to compare the corrosion protection efficiency of alkanethiol SAMs with different chain lengths and to compare SAMs with –S or –Se as their anchoring group when the samples were exposed to humidified air (HA) containing formic acid (FA). It was found that the corrosion inhibition efficiency of alkanethiol SAMs on copper was enhanced with an increased chain length. Furthermore, alkanethiol SAMs feature “selective hindrance” properties, i.e. they selectively allow the penetration of formic acid and oxygen and hinder the penetration of water resulting in the formation of copper formate and copper hydroxide. Interestingly, no formation of copper oxide was observed, which was in contrast to unprotected copper. VSFS revealed that SAMs of alkanethiols get more disordered over the course of sample exposure to HA+FA but remain attached to the substrate. Alkaneselenol SAMs initially behave similarly to their alkanethiol counterparts. However, after prolonged exposure to HA+FA they are partially removed from the surface, resulting in an accelerated corrosion of the copper substrate. Eventually, the corrosion products formed on alkaneselenol SAM covered samples consists of copper formate, copper hydroxide, and copper oxide. Figure 1

Theory-Guided Machine Learning Finds Geometric Structure-Property Relationships for Chemisorption on Subsurface Alloys

Compound flooding poses significant global socio-economic and infrastructure risks that are projected to intensify due to climate change and anthropogenic development. These compound floods, where multiple interacting drivers amplify flood extents and depths, are the most widespread and catastrophic natural hazards, in particular in urban catchments, inflicting billions of dollars in damages and jeopardizing livelihoods and critical resources. A primary challenge in addressing these events lies in the incomplete understanding of the nonlinear and complex climatic, hydrological, and hydrodynamic processes involved in compound flooding, which often leads to ineffective flood management strategies. This gap in knowledge also limits the development of suitable tools and methods for accurate flood characterization and modeling. Given the massive and escalating impacts of such events, there is a clear need for a more comprehensive understanding of the key drivers that shape flood dynamics, including uncertainties related to climate, human activity, and natural systems. Although significant advances have been made in developing physically-based dynamic models for flood simulation, these models often fall short in terms of accuracy and reliability, and remain computationally intensive for operational use. These challenges stem from an incomplete understanding of flood processes, uncertainties in predictability, and limitations in model assumptions. This presentation addresses these challenges by proposing an integrated framework that incorporates human activity, hydrological factors, topography, river morphology, and land use to enhance our understanding of riverine, coastal, and compound flood generation. It also highlights strategies for improving flood forecasting and inundation modeling through the integration of state-of-the-art process-based models, data assimilation, and machine learning, while considering cascading uncertainties in both model predictions and real-world applications.

Spatiotemporal-aware machine learning approaches for dissolved oxygen prediction in coastal waters