This presentation was held at the 2020 NAFEMS UK Conference "Inspiring Innovation through Engineering Simulation". The conference covered topics ranging from traditional FEA and CFD, to new and emerging areas including artificial intelligence, machine learning and EDA.

Resource Abstract

The paper presents the current progress of the Satratek, Novosibirsk, Russia on the creation of a digital twin of the Ansaldo Energia v64.3a gas turbine. Issues and results are presented and discussed.

At present, gas turbine plants are widely used in the Russian Federation and the rest of the world to generate electricity and heat, as well as in gas transportation. Most of these plants have a long operating time and are not serviced by the manufacturer. Also, maintenance and repair procedures are in transition from planned to condition-based.

This project will allow equipping gas turbines that are not serviced / warranted with a modular system to solve a wide range of industry tasks: early detection of emergencies, optimizing the mode of operation, predictive maintenance and repair, reducing risks of condition-based maintenance. To solve these tasks a hardware and software system is being developed.

The hardware and software system consists of the following elements: modular system for collecting and storing information; an integrated model of a gas turbine that combines the models of critical elements; data analysis and predictive analytics module; software module for solving optimization problems; graphical user interface for data visualization and issuing recommendations.

The data gathering module collects data from sensors of pressure, temperature, flow rate of fuel and air, vibration and noise installed on a gas turbine. It is responsible for conversion of signals from analog to digital, buffering, preprocessing and data transfer. Data is collected from the SCADA as well as additionally installed sensors. The data is stored for subsequent analysis.

An integrated physical and mathematical model of the turbine is deployed. It contains a model of a combustion chamber, thermal barrier tiles, blades, etc. The model is calibrated and verified by the received telemetry data.

The data analysis and predictive analytics module includes a neural network that allows detection of emergencies or suboptimal operation conditions – system imbalance, rotor curvature, large radial loads and misalignment, stator and rotor contact, loss of stability caused by leakage, cracks in the shaft, wear of blades and tiles of a turbine, etc. The module also includes a recommendation system which allows predicting the wear of components depending on the operation mode, optimizing maintenance and repair procedures.