Gas, Gas Turbines, Retrofits & Upgrades

Improved Gas Turbine Component Design via Additive Manufacturing

Issue 7 and Volume 122.

By Chris Junod, Bradley Lemke, Douglas Willham, Martin Morris, Elliott Clarke and Kevin Millen

Improved Gas Turbine Component Design via Additive Manufacturing

Utilization of additive manufacturing for gas turbine component design has been demonstrated to accelerate design cycle times, reduce development test times, provide increased test data and serve to reduce overall time to final component release. With advances in, and adaptation of, additive manufacturing, testing can now be economically incorporated as early as the concept or preliminary design phases to reduce risk of the final design configuration not meeting requirements.

Design of gas turbine compressor flow path components, especially in concept phases, has traditionally relied upon analysis-only methodologies to down-select the final design configuration. This final configuration is then screened for requirements validation near the end of the gated design reviews. It is not uncommon in these designs that phenomenon related to incorrect or incomplete boundary conditions, off design point conditions, and/or transient effects result in component performance, functionality, or life not meeting requirements.

This risk of redesign because of design requirements not being met could be reduced if testing during earlier design phases was economical and could be executed with relatively short turn times. This early testing would serve to validate analytical boundary conditions, understand and identify transient effects and ultimately help define the most meaningful final design down-select criteria. Adopting an additive manufacturing approach for prototype creation is the ideal solution for resolving the need for an economical and rapid development test results to support the growing need for acceleration of products to market. This article outlines the successfully application of additive manufacturing to aerodynamic development testing within the preliminary design phase for boundary condition definition of new compressor static flow path components for Siemens Energy

SGT-A05 industrial gas turbine engine product line. This article will review prototype design scaling and manufacturing, the methodology for aerodynamic boundary condition validation, and lessons learned from this development testing. This review will also provide a comparison of how additive manufacturing provides a cost effective and an accelerated timeframe compliant alternative to legacy development testing.

Background

Aerodynamic development and validation testing of gas turbine component designs are typically conducted in partial sector rigs, full scale component/sub-system rigs and/or in full engine test facilities. This approach to development testing always requires a compromise between capturing the perceived most important design features, rig/test availability, test facility capability, cost, and schedule. Depending upon the component being tested, the utilization of these various testing methods is usually determined by economics, thus requiring that testing be conducted in facilities that already exist that can be made to work. This method is obviously not ideal and, in many cases, highly limits potential design improvements and introduces significant risk to the final design.

Some components, like static compressor components, have traditionally been designed via analytical methods only, using a combination of one, two, and/or three dimensional analysis methods, in order to define the final design configuration from which test hardware is created. This analysis-only technique is used in all design phases before hardware is ever made available. The final designs are at best validated in a rotating rig for compressor mapping but most often are carried directly into full engine testing for validation to reduce development costs. This methodology obviously increases the amount of significant risk that the final design will not meet requirements; which a resultant failure could easily result in schedule increases of more than a year and costs which could easily double. Based on experience, the most frequent causes of these errors are poorly-defined boundary conditions (both aerodynamic and mechanical) and the discovery of other un-analyzed/quantified design drivers like transient dynamic effects.

Lessons-learned from past compressor design experience, coupled with the high paced advancement in additive manufacturing technologies, prompted Siemens Energy to adopt AM into early design aerodynamic testing for the next generation of compressor vanes. Because of the technical complexity of these new component designs, coupled with a lack of scalable and accurate legacy test data, the need for accurate inlet boundary conditions is mandatory.

Cross-Section of the Engine Flow Channel to be Characterized

Technical Approach

The model test section for this work was designed to characterize the influence of the internal flow passage and the exit plane flow coming from the zero-stage compressor

both with and without handling bleeds active. The flow passage is defined with a symmetric center-body, which represents a combination of rotor components and static end walls, and an asymmetric outer-body for bleed. All structural static internal components, e.g. struts used for sump services and bearing support, were included in the model. The test model was a geometrically similar 1/5th scale model of the stage 0 module of the engine. Focus for the test models was on the internal static flow path features only and, where possible, split line effects associated with the actual engine assembly were replicated. A representation of the engine configuration flow path is shown in Figure 1.

Printable models for the 1/5th scale test model were developed directly from the models and drawings for full-scale parts. The drawings were stripped of all details except the surfaces defining the internal flow path to expedite printing. This process left scale reproduction of the surfaces making up the center body, the outer housing, and the supporting struts. Features were then added to this flow model to add static pressure taps along key aerodynamic surfaces. Figure 2 shows a composite of the parts for the test section and how they were individually printed to mimic the engine assembly.

Within Figure 2 it should be noted that numerous non-functional cavities as well as instrumentation egress passages were included within the printed model. The concept behind open areas was to expedite printing/curing times and for instrumentation passages was to include reproducible features so that a common setup for comparable data capture could be generated for different test setups and configurations.

Parts Fabricated for the 1/5th Additive Scale Model

These parts were fabricated with a 3-D printer using a process based on stereolithography. After they were cured, they were fastened together using epoxy forming an airtight and spatially correct flow path. To improve the overall strength of the model, this assembly was then epoxied into an 8-inch diameter, standard-wall, black iron pipe section and then back-filled with a polyester resin. The resulting structure was remarkably strong. The use of this back-fill was adapted to minimize printing of non-featured items, e.g. non-flow path, because of improved material properties and because the cost of back filling was approximately 40 percent compared to the printer resin packs. The black iron pipe was fitted with pipe flanges that connected it to the air supply. A photograph of the test rig is shown in Figure 3.

The process of defining the flow path from the full-scale CAD drawings, fabricating the parts on the 3-D printer, mounting the assembly in the black iron pipe, and backfilling the voids required about two weeks and about thirty man-hours of work. The materials for each model cost approximately $1,000. Compared to traditional methods for building similar models, this approach for rig fabrication is approximately an order of magnitude faster, an order of magnitude reduction in required labor and overall components many times less expensive. As a result of this low cost and quick method, building models with interchangeable or articulated parts was not chosen since new models could be built to incorporate any changes which allowed multiple working setups available for the required flow path characterization. Swapping models into the flow system required about 30 minutes.

One exception to the part fabrication process being from polymer was the ring bounding the bleed region in the outer housing of the test section for handling bleed testing. Analysis of the ring suggested that if it were printed with the stereo lithography resin, the ring might fail. As a result, the ring was outsourced to a vendor for 3-D metal printing. Upon receipt, the metal ring was epoxied into the assembly and the more robust model was completed. Outsourcing the metal ring required an additional two weeks in the assembly of the model.

Additive Manufacturing

The majority of the parts for models used in this testing were fabricated using a Formlabs Form 2 SLA 3D printer. The printer is shown in Figure 4. The cost of the printer was approximately $3400. The print volume is 145×145×175 mm (5.7×5.7×6.9 in). The thickness of the printed layers can be chosen to be between 25 and 100 microns (0.001 – 0.004 in) with a resolution of 140 microns (0.005 in). The resin is shipped in Standard Resin Cartridges that cost approximately $150 and contain one liter of resin.

After curing, the material properties of the resin were higher than many plastics. The ultimate tensile strength is reported as 65 Mpa (9380 psia). The Young’s modulus is 2.8 GPa (402 ksi). The polyester resin that was used to backfill the printed parts and secure the parts in the black iron pipe had properties reported at about 30% higher than the printed resin. During testing there was no noticeable deformation observed or measured. Photographs of several printed parts prior to assembly are shown in Figures 5 and 6.

Testing The flow through the model was compressed air that transitioned from an inlet Mach number of about M=0.3 to a Mach number in the exit plane on the order of M=0.6. The mass flow rate tested was the dimensionless form for a compressible flow that was scaled to the mass flow rate of the actual engine section. The mass flow rate through the model was measured along with inlet stagnation pressure and temperature. The flow conditions scaling factor was defined to align most accurately with corrected flow, MrT/PA. Testing was conducted over differing corrected flows to capture effects of variable capacity with changes in inlet conditions. Testing over different corrected flows allowed for at least a partial assessment on aerodynamic conditions with changes in Reynolds numbers.

A Photograph of the Flow System Including a Scale Model Section
Formlabs Form 2, SLA 3D Printer

The operating stagnation pressure for the testing was set so the cross section Mach number immediately upstream of the mass bleed was matched to the full-scale section. The Mach number at this section was set by adjusting the upstream stagnation pressure and measuring the local static pressure. These measurements were used to calculate the section Mach number. Thirty-six static pressures were measured on the model center-body using six equally spaced rows of pressure taps downstream of the bleed section. Static pressures were also measured on the outer-body in the bleed channel, downstream of the bleed and at the exit plane.

Distortion in the exit plane was evaluated by measuring the stagnation pressure distribution. The stagnation pressures in the exit plane were measured using a rake of 30 stagnation taps located in equally spaced radial rows spanning the flow passage. The rake was also indexed radially in 15° increments to make measurements at 120 distinct locations in the exit plane annulus.

All of the data, except the pressure data, was sampled using a National Instruments Compact Remote I/O Controller containing a current input module, a temperature input module, a voltage input module, a digital module, and a current output module. The pressure data was sampled using a 16-port Scanivalve pressure scanner. Scan Rates exceeding 100 Hz were possible with the setup used with most measures being physically limited based upon fluid response times.

Testing for each test configuration could be fully setup and tested for all required data in approximately 3 hours, which included at least four compressor blow downs and total test time exceeding 10 minutes.

Examples of Model Parts Fabricated Using Stereo Lithography

Results

Testing of two different model standards over 20 different inlet conditions were completed in less than 5 months of testing. Numerous repeated tests were conducted to assess data quality, reproducibility, and if transient effects are consistent. All testing indicated highly accurate and low variability of testing, even between re-installation of models within the testing arrangement.

Results from testing have demonstrated high value to the current re-design program. Not only was the data made available during the entire preliminary design phase, working between test and engineering allowed easy utilization of the data being generated. In general, this new boundary condition data has demonstrated direct benefits to the current design, which incorporates a non-axis symmetric inlet profile feeding the high pressure compressor and also has incorporated factors which mimic the noted dynamic pressure effects associated with bleed into the design.

Bleed Ring Fabricated Using Direct Metal Laser Sintering

Conclusions

AM has been successfully applied to the preliminary design phase of a new compressor vane design within the Siemens SGT-A05 gas turbine engine product line. The results of this testing have been successfully applied and have resulted in significant improvements to the current design activity. As this testing was conducted such that benefits of AM use could be defined, the results for this current testing are:

• Rig development time was improved by roughly an order of magnitude,

• Rig and test component prototype costs reduced by more than 3x,

• Technical validity of results were only slightly lower than fully instrumented engine test results but improved relative to typical component testing,

Test results showed exceptional agreement to full scale 3D CFD results driven by rig test defined boundary conditions, and

Total Cost Comparison of current versus conventional testing was four to 10 times lower and resulted in more than an order of magnitude more data.