Skip to main content
Erschienen in: Integrating Materials and Manufacturing Innovation 4/2018

Open Access 12.11.2018 | Review Article

An ICME Framework for Incorporating Bulk Residual Stresses in Rotor Component Design

verfasst von: Vasisht Venkatesh, Ralph Green, Jaime O’Connell, Iuliana Cernatescu, Robert Goetz, Terry Wong, Brian Streich, Vikas Saraf, Mike Glavicic, Don Slavik, Rajiv Sampath, Andrew Sharp, Bill Song, Pete Bocchini

Erschienen in: Integrating Materials and Manufacturing Innovation | Ausgabe 4/2018

Aktivieren Sie unsere intelligente Suche um passende Fachinhalte oder Patente zu finden.

search-config
loading …

Abstract

Integrated Computational Materials Engineering (ICME) is an emerging discipline that aims to integrate computational materials science tools into a holistic system that can accelerate materials development, transform the engineering design optimization process, and unify design and manufacturing. A team of aerospace Original Equipment Manufacturers (OEMs) and suppliers have executed a critical program to address the United States Air Force (USAF) funded Foundational Engineering Problem (FEP) on residual stress within nickel-base superalloy components. This program was aimed at establishing methods to link predictive tools to component design functions and product realization activities with industry-wide standardized protocols. The multi-disciplinary approach links supplier and OEM materials and process models with structural analysis tools to enable manufacturing parameter selection based on disk design criteria. By linking analytical tools between the supplier and OEM, process parameters may be optimized for reduced scrap, while optimizing disk designs for design requirements. A significant challenge to doing this is qualifying and integrating sources of variation in the materials and process models with design and structural analysis tools. This paper reviews ICME infrastructure tools and methods that were used to demonstrate and validate linked residual stress-based materials and manufacturing model capabilities with design activities to achieve an optimized final component. This work was funded by the United States Air Force through the Metals Affordability Initiative (MAI).

Introduction

A team of aerospace Original Equipment Manufacturers (OEMs) and suppliers that consisted of Rocketdyne, GE, Honeywell, Rolls Royce, Pratt & Whitney, Boeing, and ATI —the Activity Integrated Product Team (AIPT)—executed a critical program to address the United States Air Force (USAF) Foundational Engineering Program (FEP) on residual stress within nickel-base superalloy components aimed at establishing methods to link predictive tools to component design functions and product realization activities with industry-wide standardized protocols. Bulk residual stresses arising from heat treatment processes can result in component distortion during subsequent machining and/or during elevated temperature service. This program focused on developing and demonstrating Integrated Computational Materials Engineering (ICME) infrastructure that incorporates bulk residual stress into manufacturing, design, and structural analysis of aeroengine disks. The use of FEPs to demonstrate the ability to systematically apply ICME throughout the aerospace supply chain and into component/systems design is of great value. FEPs provide a useful demonstration vehicle to deploy ICME tools, because tangible benefit will come directly from these efforts. Furthermore, the execution of FEP projects will enable the remaining challenges hindering full implementation of ICME tools to be identified and resolved with industry-wide methods and approaches that can be readily applied to other ICME projects.
The goal of ICME is to enable the optimization of inter-related technologies from materials, manufacturing processes, structural analysis, and component design long before components are fabricated. This will be performed by integrating the computational models involved into a holistic system. Application of ICME technology with multi-scale component design optimization is an approach to develop an enhanced materials definition that provides processing path-dependent effects and physics-based understanding of materials and subsequent component behavior. This approach for enhanced materials definition will enable greatly increased speed and accuracy, and significantly reduced cost and risk for new alloy, process, and component development and application [14].
Traditional design and manufacturing practices result in acceptable but not optimal designs with respect to cost, weight, performance, and other factors. Often costly and time-consuming iterations to material processing or design configurations are required during technology or product development programs. The use of ICME will result in closer-to-optimal designs, in fewer, more efficient iterations. The engine design and development process is expected to take less time and cost less by utilizing the ICME infrastructure to predict a component’s residual stress state. More effective use of materials will result in component weight reduction, which, in turn, will contribute to lower fuel consumption and thus lower life cycle costs. For legacy systems, ICME could allow life extensions due to enhanced understanding of a component’s residual stress state and its potential variability from design and manufacturing parameters [4]. The ability to accurately predict and control the residual stress in rotating components will result in improved engine disk designs. This is critical in engine operation due to the combined residual stress and operational stresses that can result in disk expansion affecting clearances between the rotor and engine casings. When excessive growth is observed in test engines or fielded engines, a redesign could be required to open clearances, potentially reducing efficiency of the system. This redesign, which could include a modified forging heat treatment and subsequent qualification, requires significant time and cost. The time required to approve a process change via subcomponent testing as well as engine testing can take more than a year and could cost a few million dollars if the process change requires updated material design properties. Studies have indicated various cost and time savings associated with ICME. While the actual values may vary, the conclusion is that ICME offers the gas turbine industry attractive benefits worth pursuing.
This paper reviews ICME infrastructure tools and methods that were developed under this FEP and to demonstrate and validate the capability of linking residual stress-based modeling capabilities with design activities to achieve an optimized final component. This demonstration task includes an enhanced component definition that was designed by the AIPT to demonstrate the ICME infrastructure.

ICME Infrastructure

The AIPT recognized that there are significant challenges, to full implementation of ICME, that need to be overcome within the current materials definition and design infrastructure. The team identified a number of critical ICME technical capabilities that were developed to enable a seamless, repeatable, and robust ICME infrastructure. These key ICME infrastructure capabilities are described in the following sections.

Materials Data Management Infrastructure

The purpose of the materials data management infrastructure is to securely capture all the data and information generated in this program and to allow secure access for each team member. Figure 1 depicts data flow to complete model-based component definitions at the OEM level and for the supplier to subsequently meet the model-based definitions. Types of data include quantified design requirements, CAD models, finite element process simulation inputs and outputs, material pedigree, manufacturing process parameters, material property results, component property tests, and quantified uncertainty associated with measurements and model outputs. This task not only helped effective data exchange between the AIPT members but also develops information from this task, which will have implications for ICME in general. The National Research Council (NRC), in their 2008 report on ICME, noted that one of their key tasks is to develop a “strategy for the development and maintenance of an ICME infrastructure, including databases and model integration activities” [5]. The NRC’s vision was that ICME practitioners would have “open access to a curated ICME cyberinfrastructure thereby enabling the rapid design and optimization of new materials, manufacturing process, and products.” The design of this infrastructure, therefore, was developed to not only meet the immediate needs of the project team members, but also with the larger ICME community in mind.
The infrastructure developed by the AIPT was based on using tools hosted at neutral third party sites. One such tool is iCollab, which is a managed file transfer (MFT) service managed by Aerojet Rocektdyne and hosted on a secure cloud. This MFT site was used to efficiently transfer data between team members. iCollab, was developed by Lockheed under a Defense Advanced Research Projects Agency (DARPA) initiative to securely transfer data and files between collaborators. IT security personnel from the AIPT companies validated that the security SSL protocols are in place for iCollab. Those IT security personnel also verified that communication between a user and iCollab is securely encrypted. Access to iCollab was controlled via an ID and password that was obtained from Aerojet Rocketdyne, following a background check to confirm that the applicant is a US resident and is not on any US Government restriction list.
Granta MI was chosen as the database software tool to store and share the data generated by the FEP team. A structured database like Granta MI stores data in a relational database which allows a user to search and discover the exact data required and compares it to other like data types, while also accessing the associated pedigree and metadata. ASM International (ASM) was chosen as the neutral third party to host the Granta MI database because of its commitment to serve the ICME/Materials Genome Initiative (MGI) community and willingness to allow the FEP database administer access to their network via a Virtual Private Network (VPN). ASM founded the Computational Materials Data Network (CMD Network) in order to support the MGI’s goal of having open access to digital materials data.
Every team member was only given read access to the data, while the database administrator was given both read and write access. An extra layer of security was built into the data management infrastructure by separating the development servers used for updating new datasets from the database server accessed by AIPT members. Periodically, a copy of the database on the development server was transferred to the database server that allowed team users access to updates. The infrastructure was set up this way to maximize the security of the data and to prevent any chance of accidentally exposing a company’s IP to a fellow AIPT member. While this may cause more work for the database administrator, it was the most secure approach.

Database Schema Development and Population Process

Several reports and papers have noted the significant challenge associated with unmanaged data with little metadata, limited searchability, and lacking common schema standards [13, 6, 7]. These critical data needs were addressed in this FEP. The schema of the database can be understood as the structure of the database, which is needed in order to organize the data in a manner that a user can later easily find. In a similar fashion, the data in a database needs to be organized in a structure that allows the user to be able to find data, compare the data with similar data, and understand what the data means. In addition to storing the data, all the pedigree and metadata that goes along with the data must also be organized and stored in the database. Without the metadata, the data could be useless [13]. For example, the ultimate tensile test results for a stainless steel might be 100 ksi but that data is useless without metadata information on the test parameters, conditions, and the material pedigree.
Since much of the schema and standard terminology had already been developed by the Material Data Management Consortium (MDMC), the AIPT used this as a foundation. However, additional attributes and tables were subsequently created in order to support the specific needs of the FEP datasets. A structured database, like Granta MI, was able to support the ICME and MGI goal of having searchable and discoverable digital data.
The primary goal of the data management task was to capture all the data, model files, reports, and presentations that were generated as a result of this project [6, 7]. With this as the goal, the following data types were captured [8, 9]:
1.
Chemical composition of ingot used to manufacture test articles
 
2.
Lab test reports, such as residual stress test results from the contour method
 
3.
Material specifications
 
4.
Presentations
 
5.
Best practice documents
 
6.
On-cooling tensile test data
 
7.
Residual stress model predictions
 
8.
Contour stress measurements
 
9.
Hole drill measurements
 
10.
Neutron measurements
 
11.
XRD measurements
 
12.
Quench cooling test results
 
The data management effort completed under the FEP developed an infrastructure with the ICME/MGI community in mind. It would have been easy to develop a data management infrastructure for this project alone, but the intent was to develop an infrastructure where the lessons learned could be applied by the larger ICME/MGI community.

Standard Component Specification

Incorporating location-specific material property requirements, including residual stresses, in the design of a component will require the development of new specification formats capable of conveying significantly more complicated design intent. Current material specification formats, such as AMS5662, are of the lowest common denominator type, and additional property and processing requirements often require developing completely new specifications. In general, the new specification format must be able to describe requirements for any given material property at any specific part location during any part of the manufacturing process. This hybrid specification combines the best attributes of a traditional paper specification with the location-specific details associated with digital data and models.
The component definition shall contain at least the following information;
  • Component part number and revision
  • Component description
  • Material specifications
  • Processing equipment (i.e., furnace identification) to be used, including tool numbers, where appropriate
  • Process parameters, including preparation (solvent cleaning, abrasive cleaning, etc.), furnace loading plan, time with start-of-soak definition, temperature with tolerance used for processing, atmosphere—cooling method, quenchant (if applicable), quenchant temperature with tolerance used during quenching (if applicable), quench delay (if applicable), quench time with tolerance used during quenching (if applicable), over-temperature instrument set point, time delay between consecutive heat treatment operations,
  • Property testing and/or inspection requirements, including residual stress measurements. Example location-specific properties required: tensile strength, grain size, and residual stress. All location-specific properties shall list the tolerance bands that the part must lie within upon testing. Test locations should be provided by either a diagram (for simple locations) or an embedded link (for complex test locations).
  • OEM component approval requirements, including residual stress-specific requirements. Supplier must demonstrate compliance to the entire component specification.

Material Specifications

A hybrid specification containing elements needed to develop a part-specific material specification incorporating location-based residual stress and mechanical properties was developed. This hybrid approach takes advantage of a conventional paper specification and a digital model-based specification. Such a specification would contain embedded links to 3D geometric models, model predicted 2D residual stress contours, and software-based material property files (e.g., DEFORM keyword files). The model and material property files would be located in a secure database that the supplier must log into to view and download. An example of the hybrid specification is shown in Fig. 2. The benefits of this specification format are that it combines aspects of printable specifications, definitions, legal disclaimers, and easy readability with the flexibility for the OEM to add any type of model or file that is relevant to producing the part. This could include 3D CAD files, 3D PDFs, residual stress profiles from DEFORM FEA, detailed test locations, and other applicable process specifications, e.g., heat treatments.
Location-specific material specification formats for conveying location-specific requirements include:
  • 3D models: A 3D model can be used to specify the residual stress and the mechanical properties in the part. This is the highest level of control on the part and can ultimately accommodate surface/machining and feature specific residual stress requirements.
  • Contours: 2D contour plots can be used to specify the locations and magnitude of the residual stress across a cross-section of the forging/heat treat/ship shape.
  • Zones: The part can be zoned at various levels of complexity, for example a simple zoning plan could be to divide the bore, web, and rim into specified bulk residual stress areas. An onion-skin type zone can be used to define surface residual stress requirements independently of bulk residual stresses.
  • Tolerance requirements: Every required value must have a tolerance specifying an upper and lower limit.
Requirements that should be covered in the material specification are as follows:
  • Material requirements: Can cite an existing industry/OEM standard and/or material property files for modeling.
  • Process requirements
  • Geometry requirements: Can provide CAD files, PDF, and/or geometry files for modeling.
  • Property requirements: Can include residual stress, tensile strength, grains size, etc. Can provide 3D model, 2D contour plots or files for modeling. Should include tolerances in this section of the specification.
  • Modeling requirements: Should require that the supplier perform computer modeling of the forging and heat treat process. Should also require software maturity level, model approval requirements, and acceptable output file formats.
  • Calibration and validation requirements: Should provide acceptable method for model calibration and validation, and should provide acceptable test methods and test locations. Test locations can be provided in a 3D mode, 2D drawing, or other file formats.
  • Reporting: Should provide requirements for the supplier to report data and models via the secure database.

Model Maturity Assessment and Approval

Guidelines for an OEM to assess and approve a forging suppliers’ residual stress model predictions for aerospace components was developed. Location-based residual stress requirements, with applicable tolerances, should be defined by the OEM in the Standard Component Specification. A supplier’s model, or set of models, used to predict location-based properties shall be verified, calibrated, validated, and assigned a tool maturity level (TML) that is appropriate for the OEM’s design requirements. For the FEP program’s design intent, the minimum TML level, based on the Metal Affordability Initiative (MAI) ICME TML scale, is a 4 [10]. A descriptive illustration of the ICME TML scale is presented in Fig. 3. A TML 4 required quantified understanding of the effects of processing on bulk residual stress, which the project team completed. Model verification, calibration, and validation activities were subsequently planned to meet the tool maturity level requirements for meeting the FEP design criteria.
Suppliers are responsible for model verification, validation, and testing. The OEM shall be responsible for reviewing and approving the supplier’s model results. The model results review and approval process shall cover, at a minimum, the following:
1.
Model input parameters: elastic modulus, flow stress, creep rates, boundary conditions (e.g., heat transfer coefficients, friction coefficients)
 
2.
Model verification
 
3.
Model uncertainty and sensitivity analysis
 
4.
Model validation
 

Model Verification

All models shall be appropriately verified by performing code verification and solution verification. Model verification was not rigorously pursued under the FEP as there are demonstrable examples of prior use, and documented benchmark cases for the major finite element packages, DEFORM and ANSYS.

Model Uncertainty Analysis

OEMs shall review a supplier’s methods used to quantify uncertainties in model input variables [11, 12, 13]. The steps outlined below were used to perform the uncertainty and sensitivity analysis required to understand the model.
1.
Define the model you wish to study, clearly defining the scope of the model, including pieces of code it contains in relation to one another, the inputs to the model and each code in particular, and model outputs are of primary interest.
 
2.
Define uncertainty and/or variation most pertinent to each input. This is the most important step, as it controls the output of the sensitivity analysis. This involves assigning a statistical distribution, not just a range of values, to each model input for which you wish to study the model’s sensitivity. Consider the upper and lower bounds of each distribution carefully, and use legacy data, physics and, as a last resort, expert experience, to define the distribution shape and upper and lower bounds for each input.
 
3.
Perform a full variance-based sensitivity analysis with the inputs, input ranges, and outputs defined in steps 1 and 2.
 
4.
All stakeholders for the model, including suppliers of inputs and users of outputs in design activities, should review the results of the sensitivity analysis. Results should be sanity checked to see if they meet expectations. Distributions and ranges used in the sensitivity analysis should be adjusted to reflect any discussion and decisions in this step that result
 
In addition to the sensitivity analysis, an uncertainty analysis propagates the uncertainty/variance in the inputs through the model to quantify and demonstrate uncertainty/variance in the output(s). Some software, like academic-developed shareware GEMSA, will perform an uncertainty analysis automatically when the user creates an emulator and runs a full variance-based sensitivity analysis [1416].

Model Validation and Approval

OEMs shall review experimental data used to validate and calibrate the model. Results should be analyzed using Uncertainty Quantification (UQ) methods [8]. UQ methods should be applied to determine model accuracy and range of applicability. Suppliers shall perform measurements at high sensitivity locations defined in the standard component specification. OEMs shall review and approve residual stress measurement techniques. The following are recommended techniques for location-based residual stress measurement:
1.
Strain/stress relaxation based methods (destructive): hole drilling, ring core, contour
 
2.
Diffraction-based methods where the atomic lattice serves as strain gage (non-destructive) laboratory based X-ray diffraction, synchrotron-based high energy X-ray diffraction, neutron diffraction
 
The model shall be deemed acceptable if the outputs meet all location-based residual stress requirements specified in the standard component definition. Upon acceptance of the supplier’s model, the OEM shall endorse the revision controlled model approval report.

Supplier Reporting Structure

One of the fundamental challenges of ensuring that the expectations of the design community are met is the flow of information from the supplier to the OEM. The goal of this task was to establish a framework for sharing model outputs and process information from supplier-to-OEM and supplier-to-supplier, keeping in mind that there will always be a need for IP protection. The framework aimed to define the minimum information needed that enables model output from a given process to be used as input in the subsequent process. The framework is based on a secured database, accessible by both the OEM and the supplier. For the supplier reporting structure to be successful, there must be pre-coordination between the OEM and supplier to establish non-disclosure agreements (or proprietary information agreements) and acceptable file formats for sharing. The supplier must also be given access to the secure database. Once this is in place, the OEM can easily share all material and process requirements for a given part (or set of parts) via the part-specific material specification discussed previously. The supplier can download the necessary files, perform modeling and model validation work, optimize their results, and send the model outputs to the OEM through the database. In this project, model formats have been primarily DEFORM keyword and database files. This was acceptable for the OEMs and suppliers that participated in this program. Elements of the supplier reporting structure include IP-neutral results files that are readable and displayable for both OEM and the suppliers using commercially available thermomechanical processing codes. These files include sufficient detail and data resolution to enable downstream operations and facilitate process optimization. In addition to this, the supplier reporting structure contains model predictions to the same level and at the same locations as described in the component material specification. The supplier reporting structure leverages common data protocols, standard methods, and formats developed in a previous Air Force Funded programs known as Phase II SBIR Contract: FA8650-11-C-5105, which was developed by Scientific Forming Technology Corporation and Southwest Research Institute.

ICME Workflow Demonstration

The goal of this FEP was to develop and demonstrate the ICME framework incorporating bulk residual stress into the design and manufacture of an IN718 generic rotor component (Fig. 4). The developed ICME workflow will facilitate concurrent engineering, thereby reducing design/make cycle cost and time, while also ensuring required component performance. OEMs need a secure method to convey component definition, models, and data to suppliers, while the suppliers need to provide reports back to OEM for approvals. The communication method should control and protect IP for both the OEM and the supplier, while enabling an efficient (digital) transfer of data between OEM and supplier.
There are three potential protocols for OEM/supplier interaction. In the first, the OEM performs an optimization using internal estimates for process variables and supplies a desired residual stress distribution to a forger. The forger then supplies a predicted residual stress distribution and uncertainty assessment to the OEM. In the second, the OEM would create a solid model of the desired component; the forger would create a heat treat (HT) shape, perform a HT simulation, and then furnish a residual stress distribution and uncertainty assessment to the OEM; the OEM would then perform machining modeling, structural analysis, and life assessment. In the third scenario, the OEM and forger create a common model using encrypted data to protect proprietary aspects, and subsequent optimization loops for a component would include data transfer to an FTP site. The first protocol was demonstrated in the FEP (Fig. 4); the second protocol is commonly used today, albeit with varying levels of uncertainty assessment; and the third protocol would be preferred, but not currently feasible.
OEMs and suppliers each have proprietary Intellectual Property (IP) determined by individual company business models, interests, and type of manufacturing. For example, the forger would own forge and heat treat models because they depend on company trade secrets and proprietary processing knowledge and data, whereas the OEM would own the material property data, such as forging flow stress, funded by the OEM but provided to the forger to model forging. There is the potential for joint IP that may arise during the course of business, provision of which should be made during any business agreements. In addition, both the OEM and supplier conduct business with other suppliers and OEMs, often competitors, which requires safeguards and firewalls to prevent data sharing. An OEM would only share data that it owns as IP to multiple suppliers, and a supplier would not use IP data supplied by one OEM for use in the production of another OEM’s component.

Demonstration of Model-Based Component Approval

Isight, a commercially available software system, was used to link multiple cross-disciplinary models together in an automated simulation process flow. An Isight workflow was created, linking DEFORM, a world-class software for the simulation of metal forming and heat treat operations with ANSYS, an FEA software well suited for flight cycle analysis of gas turbine engine components. The two programs were linked together so that the effect of residual stresses induced by the heat treat process on disk response in operation could be predicted. The model-based workflow takes input parameters associated with the heat treat process, simulates that process, and the subsequent material removal due to machining operations. It then performs an analysis of a high-speed spin of the final disk shape, accounting for bulk residual stresses induced by the heat treat process, and ultimately outputs heat treat residual stresses and disk growth at locations of interest. The workflow is illustrated in Fig. 5.
This workflow was used to study the sensitivity of disk residual growth to the various process input parameters and enable UQ Bayesian pre-posterior analyses to identify areas of the disk where disk growth was most sensitive to residual stress and to determine the accuracy and precision needed in the measurement at these defined locations. Due to the difficulty in measuring residual stresses accurately, this procedure enables the determination of the most appropriate measurement techniques to be applied based on available methods capabilities, penetration depth of each method, predicted strain/stress gradients, part geometry restriction, material’s microstructure, and budgetary and time constraints. Based on the measurement uncertainty expected at various locations, max/min residual stress requirements that capture part to part variation could be established.
Major elements of the developed ICME infrastructure, described previously, were executed as part of this activity. P&W and ATI demonstrated the model-based component specifications and component approval for the FEP component, by defining residual stress tolerances that balance component design requirements and supplier capabilities to deliver properties (Fig. 6). The OEM-based optimization resulted in a set of residual stress contours that define minimum and maximum allowable residual stress to ensure sufficient capability. The ATI based model optimization results were compared to P&W’s requirements in the intermediate heat teat shape condition, shown in Fig. 6. In this demonstration, the supplier capability exceeded the acceptance criterion set by the OEM.

Disk Optimization

A part of the FEP project was to assess the impact of residual manufacturing stresses on the design of IN718 engine rotors. Being able to measure and optimize residual stresses would allow for more accurate and efficient disk design. There are many ways to measure differences in design and, ideally, one would analyze life to determine an optimized design. However, because of the proprietary nature of component lifing methods among the participating companies, the AIPT was not able to compare life numbers. In order to run the optimization, boundary conditions needed to be determined. The team’s objective in setting up boundary conditions included simulating a spinning rotor with blades attached. Typically, a disk will experience radial load from centrifugal forces within the disk and from the blade mass itself. Therefore, a balance was achieved by applying a load to the disk rim to simulate the blade pull load and a rotational speed of 19,000 RPM. This combined radial force provided around 0.009 in. of permanent deformation in the bore for the baseline disk shape. In addition, sufficient load to induce plastic deformation in the part was required to be able to measure the residual growth under optimization criteria. A maximum of 0.015 in. of permanent radial growth was chosen because it was significantly larger than the manufacturing tolerances and allowed room for optimization. In addition to the radial load, another load needed to be applied to simulate the mechanical challenges of retaining a blade in the rim attachment. If only radial pull loads were applied, the optimization would lead to minimizing the rim attachment size resulting in the least amount of deflection and lightest weight. However, disk attachments must be sized to accommodate a blade. Therefore, lateral (axial) forces were added to the inside rim slot faces such that the attachment tangs would be pushed outwards. This lateral deflection would be measured as the difference in axial deformation between the top of the rim tangs, Fig. 7. This led to the second criterion for the residual axial deformation in the attachment tang (< 0.002 in.). In the FEP application, bore and rim growth on the generic disk was chosen as the critical design objective that drives decision making. These design objectives were selected for two reasons: they are very sensitive to bulk residual stresses and they could be discussed openly among the OEMs.
In order to minimize disk weight subject to constraints on disk growth accounting for the impact of bulk manufacturing residual stresses, the design targets were set to be maximum allowable plastic disk growth resulting from spinning the disk at a high rotational velocity. The growth criteria were probabilistic in nature. The probability that the radial growth would be less than 0.015 in. at the six stations illustrated in Fig. 7(a) was required to exceed 97.5% and the probability that the axial separation of the rim would increase by less than 0.002 in. Was required to exceed 97.5%, Fig. 7(b). The 97.5% criteria corresponds to a − 2 σ/50% confidence level, which is similar in intent to other aerospace requirements.
For the purposes of the FEP program, the heat treat shape was held constant over the optimization process. This was considered appropriate as the intention was to design a disk that could be manufactured and tested within the constraints of the material available to the program for machining the validation disk. The dimensions of the heat treat shape were defined to create an acceptable envelope of material for the largest allowable final machined shape given the allowable ranges of the design variables. It is recognized that the choice of a fixed heat treat shape would be unnecessarily restrictive on a clean sheet design where the heat treat supplier would be free to optimize the shape to meet OEM requirements for residual stresses in the as-delivered shape while minimizing supplier cost.
The final machined disk shape was defined by 11 key dimensions shown in Fig. 8. There are six variables in the FEP disk geometry that were chosen to be optimized and five fixed dimensions to obtain an optimum low weight design. The ranges of these dimensions were chosen to ensure a broad design space.

Disk Optimization Workflow

In order to incorporate probabilistic constraints on disk residual growth, it was necessary to define probability distributions of disk growth for each new disk design. As a result, a Monte Carlo simulation, that covered the distributions of various parameters, was used to introduce variation to the growth prediction. Conceptually, this optimization process is shown in the Isight workflow illustrated in Fig. 9.
One potential source of variation in residual disk growth is dimensional variation associated with the machining process. Typically, nominal dimensions are defined with some allowable tolerance. For this assessment, the tolerances were taken to be those to which the FEP program test disks were machined. Each dimension is represented by a normal distribution characterized by a mean (the nominal dimension) and a standard deviation. The other notable source of variation in residual growth was variability due to uncertainty in the initial residual stress distribution resulting from the heat treat process. These variations are in turn a function of the uncertainty in the various inputs to the process including processing parameters, material properties, and heat transfer coefficients. The heat treat process variables of interest to the simulation were described within ± 3σ ranges. The heat treat variables were also assumed to be normally distributed for use in the Monte Carlo simulation.
The optimization loop employed the following steps:
  • Obtain current values of disk optimization dimensions
  • Update the nominal final machined shape with the new dimensions and get the nominal design weight
  • Perform a Monte Carlo simulation on the current design using
    • –Dimensional variability based on dimensional tolerances
    • –Distributions for processing parameters
    • –Material property variability
    • –Uncertainty in heat transfer coefficients
    • –Uncertainty in axial position of the sonic shape within the heat treat shape and in the axial position of the final shape within the sonic shape to define distributions of disk residual growth
  • Extract the 97.5 percentile growth values and compare to the allowable growth criteria.
For each cycle in the Monte Carlo simulation, the following operations must be performed:
  • Geometry is updated for off-nominal dimensions
  • Solution and age heat treat processes are simulated via DEFORM
  • Material removal via machining is simulated via DEFORM
  • Disk residual growth due to a high-speed spin, and given the stress/strain state from the heat treat process, is predicted using ANSYS
In an effort to speed up run/processing times of the Monte Carlo simulation, Gaussian process emulators, which are mathematical models that stand in for, or emulate, longer runtime physics-based models, were created and used in place of the original workflow of models. These emulators, since they are essentially mathematical equations, return estimated outputs almost instantly. In this case, each emulator fit was tested to ensure that its predictions were true and sufficiently accurate to those from the original workflow. To create these emulators, an optimal Latin Hypercube design was executed with the full workflow and disk growth was extracted for each run. The results were collected and then the GEMSA software was used to build disk residual growth emulators for use in the optimization. The geometry variation was incorporated via NX modeling and simulation software and disk weight was added as an output. The process is shown in schematic form in Fig. 10.

Optimal Design

Optimal designs that met the design criteria for cases with and without bulk residual stresses were determined via the process described previously and are summarized in Fig. 11. The significant geometric differences between the nominal and optimal cases with and without residuals were located in the disk rim and at the bore-to-web fillet. This is consistent with the driving constraints for rim split and growth at the bore under the rim. A weight reduction of 11.1 lbs was achieved in the optimized disk with residual stress over the nominal design with residual stress.
The workflow described in this article was developed and executed by P&W. Each of the four OEMs successfully optimized the nominal final shape for weight and growth limits, with and without the predicted residual stress field, utilizing their internal workflow for ICME tools and methods. The optimized finished shapes evaluated by each OEM were extremely close in both size and weight. This was a culmination of successful integration of ICME between the materials and processes discipline, and the design and structural analysis disciplines.

Conclusions

The goal of the FEPs is to establish an industrial infrastructure that allows effective and repeated use of ICME tools to support design and manufacturing, while delivering tangible and meaningful benefits with the selected technology challenges. This program developed methods and demonstrated ICME infrastructure described below:
  • Materials and manufacturing process models that are validated with an approach and method that is accepted by the component/system value chain, and which include uncertainty from associated model predictions.
  • Data communication and management throughout the component/system value stream with industry standard data format structure that contains the required metadata and is extendable.
    • –Means of managing model and input data versions.
    • –Tools to transmit and upload required data into organizational data management systems.
  • Automated linkage of materials and process models with design tools
    • –Industry standard methods to establish linkages between design tools and ICME models that allow application of various industry standard tools, such as Isight.
    • –Guidelines for future ICME model development regarding methods for establishing input/output interfaces.
  • Materials definition system
    • –Development of methods and approaches to defining ICME-driven component requirements and ICME-driven material capabilities
    • –Design system approach to enable design flexibility and maximizing component capabilities with a well-defined material pedigree, i.e., component model-based requirements definition guidelines
    • –Means of communicating the material design requirements to suppliers for use in component manufacturing and delivery requirements, and for use in component quality control systems.
In addition to the ICME infrastructure elements, residual stress measurement best practices with quantified uncertainty have been developed and demonstrated. Material property test methods and DEFORM enhancements for creep models and simulating residual stress with the Bauschinger Effect were also completed.
The Uncertainty Quantification-based ICME framework developed under this program specifically assessed residual stress introduced from solution heat treatment, quenching, and subsequent aging of manufactured IN718 components. This framework should be extended to assessing location-based properties, such as microstructure, tensile strength, and creep, to improve component designs.

Acknowledgments

This work was conducted under the USAF’s Foundational Engineering Problem (FEP) on bulk residual stress development in nickel-base superalloys under the auspices of the Metals Affordability Initiative, Contract No. FA8650-13-2-5201. The support and encouragement of the FEP Program Managers (B. Song, T.J. Turner, and M.J. Caton) are gratefully acknowledged. The authors also thank Dr. Lee Semiatin at AFRL and Mr. Adrian DeWald at Hill Engineering for their excellent technical support.
This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Literatur
1.
Zurück zum Zitat Committee On Integrated Computational Engineering (2008) Integrated computational materials engineering: a transformational discipline for improved competitiveness and national security. National Academies Press, Washington, DC Committee On Integrated Computational Engineering (2008) Integrated computational materials engineering: a transformational discipline for improved competitiveness and national security. National Academies Press, Washington, DC
3.
Zurück zum Zitat Furrer DU, Dimiduk DM, Cotton JD, Ward CH (2017) Making the case for a model-based definition of engineering materials. Integr Mater Manuf Innov 6:249–263CrossRef Furrer DU, Dimiduk DM, Cotton JD, Ward CH (2017) Making the case for a model-based definition of engineering materials. Integr Mater Manuf Innov 6:249–263CrossRef
4.
Zurück zum Zitat Bolcavage A et al (2014) Integrated computational materials engineering from a gas turbine engine perspective. Integr Mater Manuf Innov 3:13CrossRef Bolcavage A et al (2014) Integrated computational materials engineering from a gas turbine engine perspective. Integr Mater Manuf Innov 3:13CrossRef
5.
Zurück zum Zitat Pollock T et al (2008) Integrated Computational Materials Engineering: a transformational discipline for improved competitiveness and national security, vol 7. the National Academies Press, Washington DC Pollock T et al (2008) Integrated Computational Materials Engineering: a transformational discipline for improved competitiveness and national security, vol 7. the National Academies Press, Washington DC
6.
Zurück zum Zitat Wong T, Venkatesh V, Turner TJ (2015) Data Infrastructure Developed for PW-8: Nickel Base Superalloy Residual Stress Foundational Engineering Problem. In: Poole W. et al. (eds) Proceedings of the 3rd World Congress on Integrated Computational Materials Engineering (ICME 2015). (Pittsburgh, PA: The Minerals, Metals & Materials Society; New York: Springer, 2015), pp. 247–259. https://doi.org/10.1007/978-3-319-48170-8_30 Wong T, Venkatesh V, Turner TJ (2015) Data Infrastructure Developed for PW-8: Nickel Base Superalloy Residual Stress Foundational Engineering Problem. In: Poole W. et al. (eds) Proceedings of the 3rd World Congress on Integrated Computational Materials Engineering (ICME 2015). (Pittsburgh, PA: The Minerals, Metals & Materials Society; New York: Springer, 2015), pp. 247–259. https://​doi.​org/​10.​1007/​978-3-319-48170-8_​30
7.
Zurück zum Zitat Ward C, Warren J, Hanisch R (2014) Making materials science and engineering data more valuable research products. Integr Mater Manuf Innov 3(1):22 Ward C, Warren J, Hanisch R (2014) Making materials science and engineering data more valuable research products. Integr Mater Manuf Innov 3(1):22
8.
Zurück zum Zitat Cernatescu I, Venkatesh V, Glanovsky JL, Landry LH, Green RN, Gynther D, Furrer DU, Turner TJ (2015) Residual-stress-measurement implementation in model-validation process as applied in the USAF Foundational-Engineering-Problem program on ICME of bulk residual stress in Ni Rotors,” 56th AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, AIAA SciTech Forum, paper AIAA–0387 Cernatescu I, Venkatesh V, Glanovsky JL, Landry LH, Green RN, Gynther D, Furrer DU, Turner TJ (2015) Residual-stress-measurement implementation in model-validation process as applied in the USAF Foundational-Engineering-Problem program on ICME of bulk residual stress in Ni Rotors,” 56th AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, AIAA SciTech Forum, paper AIAA–0387
9.
Zurück zum Zitat Rolph J, Preuss M, Iqbal N, Hofmann M, Nikov S, Hardy MC, Glavicic MG, Ramanathan R, Evans A (2012) Residual stress evolution during the manufacture of aerospace forgings, Proceedings of the 12th International Symposium on Superalloys, edited by E. Huron, R. Reed, M. Hardy, M. Mills, R. Montero, P. Portella, J. Telesman, TMS. (Warrendale, PA: TMS, 2012), pp. 881-891 Rolph J, Preuss M, Iqbal N, Hofmann M, Nikov S, Hardy MC, Glavicic MG, Ramanathan R, Evans A (2012) Residual stress evolution during the manufacture of aerospace forgings, Proceedings of the 12th International Symposium on Superalloys, edited by E. Huron, R. Reed, M. Hardy, M. Mills, R. Montero, P. Portella, J. Telesman, TMS. (Warrendale, PA: TMS, 2012), pp. 881-891
10.
Zurück zum Zitat Cowles B, Backman D, Dutton R (2015) Update to recommended practice for verification and validation of ICME methods and models for aerospace applications. Integr Mater Manuf Innov 4:2CrossRef Cowles B, Backman D, Dutton R (2015) Update to recommended practice for verification and validation of ICME methods and models for aerospace applications. Integr Mater Manuf Innov 4:2CrossRef
11.
Zurück zum Zitat Chan K, Saltelli A, Tarantola S (1997) Sensitivity analysis of model output: variance-based methods make the difference. In: Proceedings of the 29th Winter Simulation Conference, edited by; S. Andradottir, K. Healy, D. Withers, B. Nelson, IEEE Computer Society Washington, DC, USA. Chan K, Saltelli A, Tarantola S (1997) Sensitivity analysis of model output: variance-based methods make the difference. In: Proceedings of the 29th Winter Simulation Conference, edited by; S. Andradottir, K. Healy, D. Withers, B. Nelson, IEEE Computer Society Washington, DC, USA.
12.
Zurück zum Zitat Kennedy M, O'Hagan A (2001) Bayesian calibration of computer models. J R Stat Soc Ser B Stat Methodol 63(3):425–464.3CrossRef Kennedy M, O'Hagan A (2001) Bayesian calibration of computer models. J R Stat Soc Ser B Stat Methodol 63(3):425–464.3CrossRef
13.
Zurück zum Zitat Mackay DJ (1998) Introduction to Gaussian Processes. NATO ASI Series - Neural Networks and Machine Learning, Bishop (Ed.). NATO ASI Series. 168:133–165 Mackay DJ (1998) Introduction to Gaussian Processes. NATO ASI Series - Neural Networks and Machine Learning, Bishop (Ed.). NATO ASI Series. 168:133–165
15.
Zurück zum Zitat Saltelli A, Annoni P, Azzini I, Campolongo F, Ratto M, Tarantola S (2010) Variance based sensitivity analysis of model output. Design and estimator for the total sensitivity index. Comput Phys Commun 181(2):259–270CrossRef Saltelli A, Annoni P, Azzini I, Campolongo F, Ratto M, Tarantola S (2010) Variance based sensitivity analysis of model output. Design and estimator for the total sensitivity index. Comput Phys Commun 181(2):259–270CrossRef
16.
Zurück zum Zitat Reinman G, Ayer T, Davan T, Devore M, Finley S, Glanovsky J, Gray L, Hall B, Jones C, Learned A, Mesaros E, Morris R, Pinero S, Russo R, Stearns E, Teicholz M, Teslik-Welz W, Yudichak D (2012) Design for variation. Quality Engg: Special Issue: Statistical Eng. 24(2):317–345. https://doi.org/10.1080/08982112.2012.651973 Reinman G, Ayer T, Davan T, Devore M, Finley S, Glanovsky J, Gray L, Hall B, Jones C, Learned A, Mesaros E, Morris R, Pinero S, Russo R, Stearns E, Teicholz M, Teslik-Welz W, Yudichak D (2012) Design for variation. Quality Engg: Special Issue: Statistical Eng. 24(2):317–345. https://​doi.​org/​10.​1080/​08982112.​2012.​651973
Metadaten
Titel
An ICME Framework for Incorporating Bulk Residual Stresses in Rotor Component Design
verfasst von
Vasisht Venkatesh
Ralph Green
Jaime O’Connell
Iuliana Cernatescu
Robert Goetz
Terry Wong
Brian Streich
Vikas Saraf
Mike Glavicic
Don Slavik
Rajiv Sampath
Andrew Sharp
Bill Song
Pete Bocchini
Publikationsdatum
12.11.2018
Verlag
Springer International Publishing
Erschienen in
Integrating Materials and Manufacturing Innovation / Ausgabe 4/2018
Print ISSN: 2193-9764
Elektronische ISSN: 2193-9772
DOI
https://doi.org/10.1007/s40192-018-0119-6

Weitere Artikel der Ausgabe 4/2018

Integrating Materials and Manufacturing Innovation 4/2018 Zur Ausgabe

    Marktübersichten

    Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.