Gazprom Transgaz Surgut patented a method for automating the technical condition monitoring of gas pumping units. The system should convert graphs from equipment paper passports into digital form and use the obtained data to calculate GPA operating parameters.

The problem arises when working with old equipment: its passport characteristics may be presented not in tables, but as graphs. Currently, a specialist has to manually extract values from them. The patent directly calls this method labor-intensive and prone to errors.

The proposed system first scans the graph image and processes it. Then, a chatbot with generative AI determines the coordinates of the points and forms the data in the form of an XLSX table or a list of X and Y values. After that, the points are approximated by a polynomial, and the system obtains an equation describing the original curve.

Such a mathematical dependency can be loaded into an information-automated system for managing technological parameters. It compares current sensor readings with the unit's passport characteristics and calculates deviations. Based on these, it is proposed to assess the technical condition of the gas turbine unit and the centrifugal compressor.

The novelty of the solution described by Gazprom Transgaz Surgut is precisely related to the conversion of graphical passport characteristics into mathematical form using generative AI and the subsequent inclusion of these dependencies in automated calculations.

One of the scenarios is related to monitoring the flow path of the gas-air tract. A drop in air pressure after the high-pressure compressor and a change in temperature after the compressor can be used to assess the condition of the gas turbine engine. The patent proposes to take these deviations into account when determining the need for flushing the flow path.

The system also calculates the technical condition coefficients of the GTU and centrifugal compressor, takes into account fuel gas consumption, and stores the history of unit operation. The operator's screen can display current parameters, their change over time, passport dependencies, and warnings about parameters exceeding established limits.

In the example given in the patent, the use of mathematical approximation allowed obtaining a model with a coefficient of determination of 99.9%. The authors note that this approach should reduce error compared to manual reading of values from paper graphs.

The patent states that the system is designed for remote monitoring of GPA and real-time assessment of its condition. In the future, this approach should allow making decisions about the further operation of the units and their repair based on accumulated data.