The paper introduces a computer vision methodology for detecting pitting corrosion in gas pipelines. To achieve this, a dataset comprising 576,000 images of pipelines with and without pitting corrosion was curated. A custom-designed and optimized convolutional neural network (CNN) was employed for binary classification, distinguishing between corroded and non-corroded images. This CNN architecture, despite having relatively few parameters compared to existing CNN classifiers, achieved a notably high classification accuracy of 98.44%. The proposed CNN outperformed many contemporary classifiers in its efficacy. By leveraging deep learning, this approach effectively eliminates the need for manual inspection of pipelines for pitting corrosion, thus streamlining what was previously a time-consuming and cost-ineffective process.

—The paper presents the results of studying the detectability of surface volumetric defects by ultrasonic method of nondestructive testing. The possibility of using Rayleigh surface waves excited by an electromagnetic-acoustic transducer (EMAT) for these purposes is considered. Blind vertical drillings of various diameters and depths were used as artificial defects to simulate defects in low carbon steel samples. Based on the results of measurements, the dependence of the amplitude of the received signals on drilling parameters has been constructed. When statistically processing the results, the signal-to-noise ratio during the excitation of the Rayleigh wave with the help of the EMAT on defect-free areas of samples was taken into account. Optimal models of the distribution of signal amplitudes were determined to construct the curves of probability of detection (PoD) of a defect. Under the conditions of the experiments, based on the PoD curves constructed, conclusions have been drawn about the minimum dimensions of surface volumetric defects identified with 90% probability with allowance for the confidence interval of 90% and about the possibility of adjusting the parameters of ultrasonic testing by signals reflected from vertical drillings.

Issues of quality control of welded joints in hot-tool butt-welded polyethylene pipelines are considered. It is indicated that the main scope of control currently falls on inspections based on the parameters of welding and on selective mechanical testing of welded joints. Physical nondestructive post-welding testing methods are used little, which is generally not typical for the practice of testing welded joints. The reason for this is the general problem of ultrasonic methods, viz., the low detectability of contracted defects, partially transmitting ultrasound. To ensure the detectability of such defects, we evaluate the possibility of increasing the sensitivity of ultrasonic testing by using the ultrasonic time-of-flight-diffraction (TOFD) technique. It is proposed to use the tip of vertical drillings to simulate defects weakly scattering ultrasound. It is shown that when tuning is performed based on such targets, the inspection sensitivity can be increased by at least 6 dB, i.e., two times or more. In this case, the drilling tip can simulate the upper and lower edges of a planar defect vertically oriented in the section of the weld. The results of experimental studies aimed at revealing multidirectional drills with a diameter of 1 mm in welded seams of polyethylene pipes, including pipes with an additional removable outer layer, are presented. The expediency of gathering statistical data on the identification of real defects in welded butt joints of polyethylene pipes using the considered testing method is shown.

The influence of acoustic anisotropy on ultrasonic testing reliability poses a challenge in evaluating products from additive technologies (AT). This study investigates how elasticity constants of anisotropic materials affect defect signal amplitudes in AT products. Experimental measurements on AT samples were conducted to determine elasticity constants. Using Computational Modeling and Simulation Software (CIVA), simulations explored echo signal changes across ultrasound propagation directions. The parameters $A_{13}$ (the ratio between the velocities of ultrasonic transverse waves with vertical and horizontal polarizations at a 45-degree angle to the growth direction), $A_3$ (the ratio for waves at a 90-degree angle), and $A_g$ (the modulus of the difference between $A_{13}$ and $A_3$) were derived from wave velocity relationships and used to characterize acoustic anisotropy. Comparative analysis revealed a strong correlation (0.97) between the proposed anisotropy coefficient $A_g$ and the amplitude changes. Threshold values of $A_g$ were introduced to classify anisotropic materials based on observed amplitude changes in defect echo signals. In addition, a method leveraging deep learning to predict $A_g$ based on data from other anisotropy constants through genetic algorithm (GA)-optimized neural network (NN) architectures is proposed, offering an approach that can reduce the computational costs associated with calculating such constants.

The issues of classification and characterization of surface operational defects according to the results of ultrasonic, eddy current, and visual inspection methods of nondestructive testing are considered. At the same time, the visual inspection method was realized with the use of a television inspection camera equipped with a computer vision function and a laser triangulation sensor. The paper presents a dataset containing 5760 images of pipelines with and without pitting corrosion. A convolutional neural network (CNN) is presented that was applied to classify the images obtained from the TV inspection camera into images without corrosion and images with pitting corrosion. The paper presents a dataset containing 269 measurements of planar and volumetric surface defects. A model for surface defect sizing based on gradient boosting is presented. The paper develops an algorithm for classification and sizing of surface defects in complex diagnostics in which the obtained models are applied, and determines the accuracy of this algorithm in the RMSE metric, which was calculated within the studied test dataset and amounted to 0.011 mm.

In this work, a computer vision system is proposed, which allows the identification and localization of pitting corrosion on metallic surface of the gas pipelines made of low carbon and low alloy steels. For this purpose, a dataset of 5,760 images of pipeline surface with and without pitting corrosion was collected. The developed convolutional neural network (CNN) architecture was trained on this dataset. The hyperparameters of this architecture were optimized using Bayesian optimization. The developed and optimized CNN architecture does not have a large number of trainable parameters compared to existing CNN-based architectures. Also, the developed architecture showed significantly higher accuracy of 98.44 %, when classifying images into images without corrosion and with pitting corrosion. The developed CNN outperformed most existing classifiers in its parameters. A pitting corrosion localization system was also developed using the “sliding windows” and “image pyramid” methods, which made it possible to localize areas with identified pitting corrosion on the surface of pipelines made of low carbon and low alloy steels, using the developed CNN without additional labeling of the data set. The proposed deep learning approach will eliminate the need for the operator to visually inspect the pipeline for pitting corrosion, which is costly and time-consuming method. The research was carried out with financial support of the grant of Russian Scientific Fund No. 22-29-00524, https://rscf.ru/project/22-29-00524/.