Prediction and optimization of gas turbine secondary air system cooling efficiency based on deep learning

To reduce the NOx emission concentration of waste incineration boilers and improve the thermal efficiency of incinerators, the combustion process of a 600 t/d incineration boiler was numerically investigated. First, the influences of the secondary air injection angle, velocity and temperature on the NOx concentration at the waste incineration boiler outlet and the thermal efficiency of the incinerator were analyzed through a single factor simulation test. Then, coupling optimization of key operating parameters, including the secondary air injection angle, velocity and temperature, was conducted via the response surface design method to obtain the specific functional relationships between outlet NOx concentration, incinerator thermal efficiency, front wall secondary air injection angle, rear wall secondary air velocity and secondary air temperature, as well as the optimal operating parameters for the boiler. The results showed that when the secondary air injection angle of the front wall ranges from 68°∼80° and the secondary air injection angle of the back wall is 67°, the minimum NOx concentration is 142.23 mg/m3, and the maximum thermal efficiency of the incinerator reaches 85.51 %. When the secondary air velocity at the front wall is 42 m/s and the secondary air velocity at the back wall ranges from 42 ∼ 66 m/s, the NOx concentration at the outlet is the lowest at 140.05 mg/m3, and the thermal efficiency of the incinerator is the highest at 85.63 %. When the secondary air temperature ranges from 297.16 ∼ 309.16 K, the NOx concentration at the outlet is the lowest at 155.45 mg/m3, and the thermal efficiency of the incinerator is the highest at 84.64 %. The secondary air injection angle, velocity and temperature impose significant effects on the NOx concentration at the outlet and thermal efficiency of the incinerator. The optimal parameters, as determined in the multifactor simulation test, include a 77° secondary air injection angle of the front wall, 69 m/s secondary air velocity at the back wall, and 297.15 K secondary air temperature. Under these conditions, the NOx concentration at the outlet is 134.98 mg/m3, and the thermal efficiency of the incinerator reaches 86.11 %. This study has important guiding significance for reducing pollution and improving the efficiency of waste incineration boilers.

Conventional gas turbine secondary air system in early stage typically uses a fixed throttle unit between supply side which is compressor bleeding point and demand side which is turbine blade. The cooling air mass flow strongly depends on the extraction pressure characteristics of compressor. Optimal amount of cooling air is supplied only in design point in this way. The cooling air mass flow would be either too much or too less in off design condition. Recently, heavy duty gas turbine manufacturers introduced an active control method for secondary air system. The main strategy is to adjust the cooling air valve set point as a function of gas turbine load percentage in order to adjust cooling air pressure ratio and cooling air mass flow as well. With this active control strategy, cooling mass flow is separated from compressor extraction pressure characteristics, and it can provide a better way to deal with combustion contaminant issues. But it is still a problem that there is no dependence relationship between cooling air valve set point and operating ambient temperature in that strategy. That is to say, the cooling air pressure ratio is constant while varying ambient temperature at base load. In order to quantitatively analyze this phenomena, a 1-dimensional integrated gas turbine thermodynamic analysis method is first applied to obtain the extraction pressure characteristics of compressor for all bleeding points. In the meantime, the optimal cooling air mass flow for turbine blades in different operating conditions is evaluated by a 0-dimensional heat transfer assessment method. A 1-dimensional fluid network analysis method is then employed to calculate the cooling air mass flow variation characteristics for 2 typical throttle configurations between compressor bleeding points and turbine blades, the first one is setting a fixed throttle unit, and the second one is setting constant cooling air pressure ratio by a cooling air control valve. Quantitative calculation results show that the cooling air supply will not always meet the optimal requirements at different ambient temperature conditions with neither of the 2 configurations. This paper further optimized the active control strategy. With the optimized strategy, cooling air supply not only no longer depends on extraction characteristics of compressor, but also could be actively adjusted according to the optimal requirements of turbine blades at different ambient temperature conditions. Performance evaluation results show that the optimized active control strategy could enhance the overall efficiency without exceeding maximum allowable metal temperature of turbine blades.

The Secondary Air System (SAS) of a jet engine is an open system: there is a bleed off take from the compressor — at the lowest possible pressure compatible with the sink where the flow is to be discharged —, the air then travels through the internal cavities of the engine cooling down the compressor and turbine discs and sealing and cooling bearing chambers. Eventually, the air is discharged at the turbine rims preventing the air of the main gas path from entering the internal turbomachinery cavities that would damage the turbine assembly. Ultimately the system is also primarily responsible for determining the endloads exerted on the turbine discs. The amount of air bled from the main gas path, although necessary, impairs the engine performance because it is purged from the main engine cycle. In order to quantify and minimise its pernicious effect, the usual practice is to model the engine SAS in steady state conditions with 1D network solvers where the net nodes represent the various components of the system. For usual engine transients it is sufficient to analyse the system performance with a quasi steady approach because the time constant of the air system is insignificant compared with the turbomachinery characteristic time. Nonetheless, the rapid changes that occur during slam accelerations or failure scenarios — particularly shaft failure events — call for a different approach to calculate the endloads fluctuations. However, to the author’s knowledge, there is not such an approach to predict the transient response of the system available in the literature hitherto. The aim of the present research is to develop a dynamic model for gas turbine secondary air systems capable of tackling the sudden changes in the flow properties that occur within the system in the aforementioned cases. The whole system is initially broken down into a series of chambers of a finite volume connected by pipes that are initially modelled in isolation and then interconnected. The resultant tool constitutes a baseline onto which further improvements and modifications will be implemented in subsequent works. This first part of the paper explains the mathematical apparatus behind the model of the two main components of the SAS — chambers and pipes — isolated. Then the assumptions made and the limitations that arise as a result are described thoroughly. Finally, the computational results obtained are successfully compared against experimental data available in the public literature for validation.

In order to improve performance of heavy-duty gas turbines, in terms of efficiency and reliability, accurate calculation tools are required to simulate the SAS (Secondary Air System) and estimate the minimum amount of cooling and sealing air to ensure the integrity of hot gas path components. A critical component of this system is the cavity formed between coaxial rotating and stationary discs, that needs a sealing flow to prevent the hot gas ingestion. This paper gives a general overview of a 1D tool for the analysis of stator-rotor cavities and its integration into an “in-house” developed fluid network solver to analyse the behaviour of the secondary air system over different operating conditions. The 1D cavity solver calculates swirl, pressure and temperature profiles along the cavity radius. Thanks to its integration into the SAS code, the cavity solver allows estimation of sealing air flows, taking into account directly of the interaction between inner and outer extraction lines of blades and vanes. This procedure has been applied to the AE94.3A secondary air system and the results are presented in terms of sealing flows variation for the cavities of second and third vane on gas turbine load and ambient conditions. In some different load conditions, calculated secondary air flows are compared to experimental data coming from the AE94.3A Ansaldo fleet.

One of the most critical parts of a modern gas turbine that its reliability and performance has a great influence on cycle efficiency is the secondary air system (SAS). Modern systems functions to supply not only cooling air flow for turbine blades and vanes but sealing flow for bearing chambers and turbine segments as well as turbine disks’ purge flow in order to eliminate hot gas ingestion. Due to the various interactions between SAS and main gas, consideration of the former is substantially crucial in design and analysis of the whole engine. Geometrical complexities and centrifugal effects of rotating blades and disks, however, make the flow field and heat transfer of the problem so complicated AND too computationally costly to be simulated utilizing full 3-D CFD methods. Therefore, developing 1-D and 0-D tools applying network methods are of great interests. The present article describes a modular SAS analysis tool that is consisted of a network of elements and nodes. Each flow branch of a whole engine SAS network is substituted with an element and then, various branches (elements) intersect with each other just at their end nodes. These elements which might include some typical components such as labyrinth seals, orifices, stationary/rotating pipes, pre-swirls, and rim-seals, are generally articulated with characteristic curves that are extracted from high fidelity CFD modeling using commercial software such as Flowmaster or ANSYS-CFX. Having these curves, an algorithm is developed to calculate flow parameters at nodes with the aid of iterative methods. The procedure is based on three main innovative ideas. The first one is related to the network construction by defining a connectivity matrix which could be applied to any arbitrary network such as hydraulic or lubrication networks. In the second one, off-design SAS calculation will be proposed by introducing some SAS elements that their characteristic non-dimensional curves are influenced by their inlet total pressure. The last novelty is the integration of the blades coolant calculation process that incorporates external heat transfer calculation, structural conduction and coolant side modeling with SAS network simulation. Finally, SAS simulation of an industrial gas turbine is presented to illustrate capabilities of the presented tool in design point and off-design conditions.

This paper presents a novel approach that couples system modelling of both the thermo-fluid system and the thermal solid system by modelling conjugate heat transfer within a single 1D system analysis solver and applies it to the Secondary Air System of a Gas Turbine. The Secondary Air System design has to balance minimizing engine bleed whilst ensuring sufficient cooling. To achieve this, designers model both the secondary air flow and the temperature distribution in solid components. System CFD tools like Simcenter Flomaster may be used to solve flow, pressure and temperature distributions and a 3D thermal solver used to perform the thermal analysis of the blade and disc solids. The thermal interaction between the secondary flow system and the solid components is a key part of the model and is known as conjugate heat transfer analysis (CHT). This approach is problematic early in the design cycle when detailed or stable geometry information may not be available for the 3D thermal tool. An approach that couples the modelling of both the thermo-fluid system and the thermal solid system within a single 1D/system analysis tool offers the advantage of faster modelling and consistent model accuracy of both fluid and solid components, especially in the early concept design stage. This 1D-CHT approach has been implemented within Simcenter Flomaster and validated using an idealized analytical solution. It is shown that the model can be applied to the analysis of gas turbine secondary air systems including cavity flows and thermal analysis of the rotor and stator discs that form the thermal boundary of these cavities using Simcenter Flomaster alone.

Transient modelling and simulation of gas turbine secondary air system

Hot gas ingestion from the main gas path into the wheelspace of gas turbines can lead to severe thermal damage to critical components, reducing operational reliability and lifespan. Hence, predicting and mitigating hot gas ingestion is crucial for improving turbine efficiency and durability. This study conducts a novel and comprehensive 3D numerical analysis to investigate hot gas ingestion and rim seal effectiveness in the first wheelspace of the Frame 9 gas turbine at Shahid Rajaei Power Plant (Qazvin, Iran). Novelty of this work lies in its use of a high-fidelity, 3D computational Fluid Dynamics (CFD) model that includes actual engine operating conditions, first-stage vane and blade geometries, and all relevant cooling paths (vane convection, film cooling, and rim cavity leakage), which allowed for a detailed investigation of the complex fluid dynamics often oversimplified in conventional models. Furthermore, this study introduces an integrated methodology that couples the CFD results with a 0D Secondary Air System (SAS) code to model the system-level thermal effects of ingestion. The analysis revealed several key conclusions. The effectiveness curve of the rim seal, extracted under these realistic conditions, diverges significantly from conventional effectiveness curves under off-design conditions. A significant result of the integrated analysis is the establishment of a fundamental relationship between the dimensionless ingestion and sealing parameters. Results reveal that maintaining the sealing parameter above a certain threshold is crucial for mitigating hot gas ingestion, which offers a direct and actionable guideline for optimizing sealing flows. This research presents a robust methodology and provides important insights into the performance of the F9 turbine’s secondary air system, offering valuable guidance for optimizing sealing flow rates to enhance turbine longevity and efficiency.

The dependable design of secondary air system is one of the main tasks for the safety, reliability and performance of gas turbine engines. To meet the increasing demands of gas turbine design, improved tools in prediction of the secondary air system behaviour over a wide range of operating conditions are needed. A real gas turbine secondary air system includes several components, therefore its analysis is not carried out using a complete CFD approach. Usually, these predictions are performed using codes, based on simplified approach which allows to evaluate the flow characteristics in each branch of the air system requiring very poor computational resources and little calculation time. Generally the available simplified commercial packages allow to correctly solve only some of the components of a real air system and often the elements with a more complex flow structure cannot be studied; among such elements, the analysis of rotating cavities is very hard. This paper deals with a design-tool developed at the University of Florence for the simulation of rotating cavities. This simplified in-house code solves the governing equations for a steady one-dimensional axysimmetric flow using experimental correlations both to incorporate flow phenomena caused by multidimensional effects, like heat transfer and flow field losses, and to evaluate the circumferential component of velocity. The simplified calculation tool was designed to simulate the flow in a rotating cavity with radial outflow both with a Batchelor and/or Stewartson flow structures. Several studies have been carried out by the authors to develop suitable correlations for the discs friction coefficients and for co-rotation factor evaluation. The results of these analyses are available in the literature. In the present paper the authors develop, using CFD tools, reliable correlation for rotor disk pumped mass flow rate and provide a full 1D-code validation comparing, due to a lack of experimental data, the in-house design code predictions with those evaluated by CFD.

As temperatures in jet engine combustion chamber and turbine increase, the relevance of heat transfer between structure and secondary airflow is increasing. Secondary air properties greatly influence material temperatures, which control the thermal expansion of engine parts. Redefining tip clearances and seal gaps, this modifies considerably pressure losses and mass flow rates in the air system, impacting flow and material temperatures. The coupling effect, estimated strong, must be accurately addressed to obtain an optimized secondary air system design. The aerodynamic calculations yielding mass flows and pressures, and thermal analysis providing temperatures and the material expansion are performed separately. In reality, there are interactions currently not considered which would require a lot of time-consuming iterations. The present investigation aims at taking this mutual interaction into consideration in a robust and modular analysis tool, combining secondary air system, thermal and mechanical analysis. For air system calculations, a network representation with nodes, chambers, connected by pressure loss devices is used. This network is coupled with a thermomechanical model of the engine secondary air system in the free finite element software CalculiX® [1]. This paper presents the first module implemented in CalculiX®, dedicated to solve typical Secondary Air System problems and gives an insight on the implementation of the coupling process currently under development.

The rise in temperature and pressure ratios for both turbine and compressor intrinsically require an increased secondary air system performance. The more demanding operating environment of the components to be cooled and the quest for improved efficiency, command a particular attention to the losses faced along the secondary flow path. One of the major pressure drop is experienced with the radial inflow in a rotating cavity, becoming a potential high gain branch of the system. The use of vortex reducers has shown a considerable improvement in performance and this paper presents the work done to minimize the pressure losses for an industrial gas turbine application using various vortex reducers. The numerical results and the set-up of a one-dimensional network model and Computational Fluid Dynamics (CFD) model for various configurations are discussed in detail showing an improved system performance with the proposed vortex reducer configuration.

It is important to monitor the quality of the air used in the cooling system of a gas turbine engine. There can be many reasons that particulates smaller than the minimum size removed by typical engine air filters can enter the secondary air system piping in a gas turbine engine system. Siemens has developed a system that provide real time monitoring of particulate concentrations by adapting a commercial electrodynamic devise for use within the confines of the gas turbine secondary air system with provision for a grab sample option to collect samples for laboratory analysis. This on-line monitoring system is functional at typical engine cooling system piping operating pressure and temperature. The system is calibrated for detection of iron oxide particles in the 1 to 100 micrometer range at concentration of from 1 to 50 parts per million mass wet (ppmmw) The electro dynamic device is nominally operable at 800°C. The particulate monitoring system requires special mounting and antenna. This system may be adjusted for other materials, sizes and concentrations. The system and its developmental application are described. The system has been tested and test results are reviewed. The test application was the cooling air piping of a Siemens gas turbine engine. Multiple locations were monitored. The cooling system in this engine incorporates an air cooler and the particulate monitoring system was tested upstream and downstream of the air cooler for temperature contrast. The monitor itself is limited to the piping system and not the engine gas-path.

Reliable design of secondary air system is one of the main tasks for the safety, unfailing and performance of gas turbine engines. To meet the increasing demands of gas turbines design, improved tools in prediction of the secondary air system behavior over a wide range of operating conditions are needed. A real gas turbine secondary air system includes several components, therefore its analysis is not carried out through a complete CFD approach. Usually, that predictions are performed using codes, based on simplified approach which allows to evaluate the flow characteristics in each branch of the air system requiring very poor computational resources and few calculation time. Generally the available simplified commercial packages allow to correctly solve only some of the components of a real air system and often the elements with a more complex flow structure cannot be studied; among such elements, the analysis of rotating cavities is very hard. This paper deals with a design-tool developed at the University of Florence for the simulation of rotating cavities. This simplified in-house code solves the governing equations for steady one-dimensional axysimmetric flow using experimental correlations both to incorporate flow phenomena caused by multidimensional effects, like heat transfer and flow field losses, and to evaluate the circumferential component of velocity. Although this calculation approach does not enable a correct modeling of the turbulent flow within a wheel space cavity, the authors tried to create an accurate model taking into account the effects of inner and outer flow extraction, rotor and stator drag, leakages, injection momentum and, finally, the shroud/rim seal effects on cavity ingestion. The simplified calculation tool was designed to simulate the flow in a rotating cavity with radial outflow both with a Batchelor and/or Stewartson flow structures. A primary 1D-code testing campaign is available in the literature [1]. In the present paper the authors develop, using CFD tools, reliable correlations for both stator and rotor friction coefficients and provide a full 1D-code validation comparing, due to lack of experimental data, the in house design-code predictions with those evaluated by CFD.

In the first part of this paper the equations and results for the transient models developed for the SAS components in isolation have been thoroughly explained together with the assumptions made and the limitations that arose subsequently. This second part explains the work carried out to couple the individual components into a single network with the aim of assembling a dynamic model for the whole engine air system. To the authors’ knowledge the models published hitherto are only valid for steady or quasi steady state. It is then the case that the differential equations that govern the fluid movement are not time discretised and thus can be solved in a relatively straightforward fashion. Unlike during transients, the flow is not supposed to reach sonic conditions anywhere within the network and most important, flow reversal cannot be accounted for. This study deals with the mathematical apparatus utilised and the difficulties found to integrate the single components into a network to predict the transient operation of the air system. The flow regime — subsonic or supersonic — and its direction have deemed the choice of the appropriate numerical and physical boundary conditions at the components’ interface for each time step particularly important. The integration is successfully validated against a known numerical benchmark — the De Haller test. A parametric analysis is then carried out to assess the effect of the length of the pipes that connect the system cavities on the pressure evolution in a downstream reservoir. Transient flow through connecting pipes is dependent on the fluid inertia and so it takes a certain time for the information to be transported from one end of the duct to the other. As it would be expected, the system with a longer pipe is found to have a longer settling time. Finally, the work concludes with the analysis of the flow evolution in the secondary air system during a shaft failure event. This work is intended to continue to address the limitations imposed by some of the assumptions made for an extended and more accurate applicability of the tool.

In this paper probabilistic methods are applied to a 1D flow model of the Secondary Air System (SAS) of an industrial gas turbine. An overview of the methods applied and the results which can be provided by the probabilistic analysis is given. The paper especially deals with the problems that arise from the high number of probabilistic input parameters connected with a simulation of the Secondary Air System. To overcome these problems a fast method for the detection of nonlinear dependencies is introduced. For improvement in the development process a numerical test plan of the SAS is used to calculate response surfaces describing the system. This allows improvement of the SAS using the response surfaces instead of Monte Carlo Simulations and therefore results in a significant reduction of required flow calculations during the improvement process towards a more robust design. Furthermore this provides a possibility to judge the effects of changes in the input parameters without additional flow calculations.