ru / en

“PETON” Implementing Artificial Intelligence Technologies

2021-08-16

Digital transformation is giving rise to a new reality for the oil and gas industry, and engineering companies are heavily involved in the digitalisation process.

In 2020, the company launched a digital transformation programme that provides for IT improvement and digital transformation of business. The company has launched a number of AI-related pilots along with Process Mining and data analytics projects.

“Large-scale projects implemented by “PETON” require the launch of cutting-edge technologies, automation of business, and digital transformation”, said the representatives of SRDI O&G “PETON”.

“PETON” has implemented a defect detection automation project using AI technologies.

The problem of detecting and minimising the adverse consequences of equipment defects is quite acute in the construction industry. “PETON” has implemented AI technologies (Computer Vision) for early detection and risk management in this area.

Traditional defect detection requires a significant labour input at the stages of procurement and acceptance traditionally associated with the department of supplies. Procurement in the oil and gas sector involves a large number of items, and each article may prove to contain a potential defect that has to be detected at the incoming control stage. Besides, potential defects may emerge upon installation. 

Implementation of digital technologies provides a number of economic and managerial benefits: instead of involving several high-end employees in defect detection at the construction facility, it is enough to appoint a single rank-and-file employee with a lower qualification. In addition, early detection reduces the equipment replacement and transportation costs significantly. 

“Automation of defect detection reduces the company’s expenses, ensures a higher quality of works, and makes it possible to manage the risks that may transform into adverse consequences tomorrow”, said the representatives of “PETON”.

YOLO v4 neural network has been picked up to detect defects on pictures. According to paperswithcode.com, this is currently the fastest neural network to detect items in video streams. The preliminary stage provided for collection of more than 2000 pictures required for training where each class of defects was marked. Later on, the training set of pictures was expanded for the neural network to learn to understand that an item of the same class may look differently and that despite the fact that the image quality depends on a number of factors in real life, the neural network is supposed to produce a correct result in various conditions. 

To verify the defect, the construction site operator will create a mail on the smartphone, attach the pictures of the valves thereto, and send it to the corporate e-mail. A chat bot will process the letters, launch the neural network to detect defects, and return the result. In case of defects, a message with the defects revealed and a report will be delivered to the controller by default in order to make the removal decision.

Thus, SRDI O&G “PETON” proved able to automate defect detection by means of a neural network. In the near future, the project design provides for the expansion of the list of articles for defect detection, increase in the number of the classes of defects that may be detected, and development of a special app to facilitate the management of the neural network. 

The company takes energetic efforts to implement AI technologies into its operations and is planning to launch a number of digitalisation projects by the end of 2021.