Software testing CyberStudio software product for predictive analytics of gas turbines GTES-25PA
MACHINERY
Objective: Exploring the possibility of using predictive models for equipment
Tasks: Creation of a mathematical model for diagnostics of technical condition of gas turbine engines as part of the GTES-25PA in a specialized CyberStudio software.
Equipment: 4 gas turbine engines
Solution
Analysis of historical data of equipment operation and building models of operation of 4 gas turbine engines in a specialized CyberStudio software.
Results
The models detect a number of anomalies in the data associated with certain works performed on engines, which indicate the development of malfunctions, about the possible change in the characteristics of the system.
This is how the discrepancy is observed in the following models:
Tasks: Creation of a mathematical model for diagnostics of technical condition of gas turbine engines as part of the GTES-25PA in a specialized CyberStudio software.
Equipment: 4 gas turbine engines
Solution
Analysis of historical data of equipment operation and building models of operation of 4 gas turbine engines in a specialized CyberStudio software.
Results
The models detect a number of anomalies in the data associated with certain works performed on engines, which indicate the development of malfunctions, about the possible change in the characteristics of the system.
This is how the discrepancy is observed in the following models:
- Increases between flushes discrepancy in HEU filter models
- Models respond to non-critical node malfunctions
Development of a system for marine gas turbine engine diagnostics using digital twin
MACHINERY
Objective: Develop a system for diagnostics of marine gas turbine engines using digital twin technology
Tasks:
Solution
CyberStudio software for creation of mathematical models of marine gas turbine subsystems in a proper working condition and setting a defect classifier
Results
Diagnostic models for marine gas turbine engines have been developed:
1) A flow path model based on a hybrid physical-neural network architecture, identified from test data.
2) A fuel system model based on a fully connected neural network architecture, trained on external physics-mathematics based models and fine-tuned on test data.
3) An oil system model based on a neural network architecture with long short-term memory, trained on test data.
Simulation of gas turbine engine systems with defective conditions was conducted, and a classifier for gas turbine engine defects was created:
1) For the flow path, more than 10 defects were identified using the created hybrid flow path model.
2) For the fuel and oil systems, more than 10 defects were identified using physics-based modeling in an external program.
3) All simulated defects were collected in a defect state archive, and a defect classifier was built based on it.
A prototype of a predictive diagnostic system for marine gas turbine engines has been created.
The correctness of the diagnostic algorithms was tested on test data from a gas turbine engine with simulated defective conditions
Tasks:
- Development of high-precision physical and mathematical models of oil, fuel and air subsystems of gas turbines
- Identification of developed models by data test results
- Creation of data results archive
- Defect classifier configuration
Solution
CyberStudio software for creation of mathematical models of marine gas turbine subsystems in a proper working condition and setting a defect classifier
Results
Diagnostic models for marine gas turbine engines have been developed:
1) A flow path model based on a hybrid physical-neural network architecture, identified from test data.
2) A fuel system model based on a fully connected neural network architecture, trained on external physics-mathematics based models and fine-tuned on test data.
3) An oil system model based on a neural network architecture with long short-term memory, trained on test data.
Simulation of gas turbine engine systems with defective conditions was conducted, and a classifier for gas turbine engine defects was created:
1) For the flow path, more than 10 defects were identified using the created hybrid flow path model.
2) For the fuel and oil systems, more than 10 defects were identified using physics-based modeling in an external program.
3) All simulated defects were collected in a defect state archive, and a defect classifier was built based on it.
A prototype of a predictive diagnostic system for marine gas turbine engines has been created.
The correctness of the diagnostic algorithms was tested on test data from a gas turbine engine with simulated defective conditions
Intelligent monitoring and predictive modeling of the technical condition of the steam turbine and turbogenerator
MACHINERY
Objective: Development of a software package for intelligent monitoring and predictive modeling of the technical condition of the steam turbine and turbogenerator
Tasks:
Solution
Development of steam turbine and turbogenerator models in CyberStudio
Results
Software package has been developed
A set of documentation has been developed in accordance with the technical task
The software package has been connected to real data and put into operation
Tasks:
- Exploratory statistical analysis of turbine and turbogenerator characteristics, development of a technical task for creating a software package for intelligent monitoring and forecasting of turbine and turbogenerator parameters, development of prototypes of the main modules.
- Development of a software package for intelligent monitoring and forecasting of turbine and turbogenerator parameters, retraining of data analysis models, debugging in accordance with the acceptance test protocol
- Steam turbine
- Turbogenerator
Solution
Development of steam turbine and turbogenerator models in CyberStudio
Results
Software package has been developed
- Software components for data loading and preparation
- Components for setting up and calculating data analysis models
- Control module for computing and data exchange
- Database for storing equipment monitoring data
- Steam turbine
- Turbogenerator
A set of documentation has been developed in accordance with the technical task
The software package has been connected to real data and put into operation
Optimization of the pressure injection molding process for plastic parts
MACHINERY
Objective: Optimize the pressure injection molding process for plastic parts
Tasks:
Equipment: Plastic parts
Solution
Tasks:
- Reduce resource consumption in the production of plastic parts
- Increase the speed and efficiency of technical processes
- Accelerate the process of selecting and predicting the optimal molding conditions
Equipment: Plastic parts
Solution
- Creation of a digital model of the pressure injection molding process for plastic
- Deep learning technology utilization on over 200 parts
- Knowledge combination of the physics of plastic flow processes with the client's expertise
- Potential for scaling up to €500,000
- Prediction of the optimal molding conditions in 2 seconds instead of 20 days
- Optimization of the placement of injectors and recommendations for modifying conditions
