Towards a grid enabled Engineering Body Scanner

advertisement

Towards a grid enabled Engineering Body Scanner

Kevin T. W. Tan

1

, Greg Johnson

1

, Nick J. Avis

2

, Philip J. Withers

1

1

Manchester Materials Science, University of Manchester, Grosvenor Street,

Manchester M1 7HS, UK

2

School of Computer Science, Cardiff University, Queen’s Buildings, Newport Road,

P.O. Box 916, Cardiff CF24 3XF, UK

Abstract

1.

We outline an integrated information management system employing ontology-based knowledge integration within a XML-based text/data master file. This is to assist material scientists to systematically acquire residual stress and damage measurements, register and co-visualise these datasets from components of complex geometry. At the present time, 3D residual stress and damage imaging experiments can be very complicated to undertake and manage for complex shape component. A great deal of valuable measurement effort and time is lost in setting up the experiment. Instrument visualisation systems have been developed to import component geometry data and the physical configuration for different strain scanning instrument environments to assist with pre-measurement planning. Our system builds on this work with the aim of providing an XML knowledge-based system to assist the materials scientist in the systematic gathering and collation of data. The integrated XML-based database system provides a framework for pre- and postmeasurement data collection and storage. It also acts as a transparent and distributed data management tool for the registration, fusion and visualisation of different collected datasets as well as comparison with 3D predictions (finite element) computer models. Potential of mounting current system to become grid-enabled experiments with distributed visualisation can benefit the residual stress experiments to allow geographically distributed material scientists contribute their knowledge towards different part of the experimental samples.

Background

To predict the lifetime of a component, it is not sufficient to know only the geometry and the stresses externally applied on it. The lifetime is usually determined by the interactions between the defects within the component and the stresses to which it is exposed. These stresses are a combination of those applied in service and those which pre-exist within the component as a consequence of manufacture and previous service, namely the residual stresses. Residual stresses in a structural material or component are those stresses which reside in the object without the application of any service or other external loads. Manufacturing and fabrication processes such as casting, welding, machining, moulding, heat treatment, etc. all normally introduce residual stresses into the manufactured object. Another frequent cause of residual stress is in-service repair or modification. In some cases, stress may also be induced later in the life of the structure by installation or assembly processes or by occasional overloads.

Depending upon the magnitude, sign, and distribution of the stress with respect to the load-induced stresses, the effects of residual stress may be either beneficial or detrimental.

The residual stresses are very commonly detrimental, and there are many documented cases in which these stresses were the predominant cause contributing to unexpected fatigue and other structural failures. This is because the service stresses were superimposed on residual stresses that were greater than expected. The particularly insidious factor of residual stress is that its existence generally goes unpredicted until after malfunction or failure occurs. The presence of damage or defects can also compromise a component’s performance to give unexpected failures. Thus, it is important to be able to map residual stress and damage if life-times are to be predicted accurately.

Whilst applied stresses can be accounted for in the design of a component, it is more difficult with residual stresses because they are difficult to predict and measure reliably. Residual strain measurement is now possible at high spatial resolution non-destructively at depth so as to

Figure 1: Wide-chord Fan Blade characterised using a suite of different instruments build up 3D maps, but has its own severe limitations. Most of the residual stress and to mount the sample a number of times, either because of different regions of interest, or to damage measurement methodologies have been developed on flat or cylindrical test-specimens, and are not readily adaptable to real test objects of arbitrary size and shape. In fact a recent compare measurements made in different directions, or on different instruments. For example, consider the wide-chord fan blade shown in Figure 1 , there are at least two regions round robin study looking at the magnitude and distribution of residual stress in an Al ring and plug test sample, found that while there was good agreement in the level of residual strains measured, there was a wide variation (standard deviation 0.2mm) in the ability to locate the sample [1] . Clearly an inability to measure at the right location can be very serious in of interest for measurement: i) the root and ii) the fan blade body. Finally it may be necessary to apply and combine the results from various measurement techniques, such as neutron, synchrotron and lab X-ray diffraction for stress measurement and X-ray tomography for internal structural observation.

As the components that are of interest to the locations were the gradient in strain or stress is steep. For example at a peened surface, gradients as steep as 2000MPa/mm are not uncommon, in which case a positioning error of

0.2mm equates to 400MPa uncertainty in stress.

Poor alignment of simple rectangular samples can lead to significant errors in the position of measurement, for example a mounting misalignment or sample distortion of just 1° for a 10cm long component can result in location errors of ~2mm. This problem is exacerbated for complex samples, especially when fiducial marks are hard to define. Furthermore, the problem may become compounded by the need engineering community become more complex in shape, an increasingly large amount of valuable measurement effort and time is lost in setting up an experiment on a particular instrument. At the same time, the point measurement times have fallen, meaning that the long set-up times have become the major obstacle to increasing through-put. In addition, since each measurement instrument has different measurement geometries, sample holders, etc. it is not usually possible to save time by re-using established configurations. At the present time, scans are undertaken using the laboratory frame specific to the particular

Figure 2: System architecture for the existing engineering body scanner instrument and sample orientation.

Consequently, the co-registration of data from different scans or instruments is very difficult.

Furthermore, the component may need to be measurements, and register, process and covisualize these datasets. Further we examine how the existing single integrated system could be incorporated within a grid-based resource and knowledge-sharing infrastructure. removed and reoriented several times to complete a set of measurements. The problem is even worse when data sets must be combined, for example when strain measurements made in three perpendicular directions must be merged to calculate stresses. Even small registry errors can combine to give large errors in stress measurement. Combining the data is often a major undertaking simply because different coordinate systems have been used for the different sets. This is because the sample must be removed and reoriented to facilitate their successive measurement. Furthermore, opportunities for bringing together results from different measurement techniques are often missed because information is stored in different formats, on separate software using different co-ordinate systems.

This paper describes an integrated system employing ontology-based knowledge integration and an XML-based text/data master file to assist materials scientists to acquire systematically for components of complex geometry geometric, residual stress and damage

2.

Existing System Architecture

The Engineering Body Scanner (EBS) project has targeted two challenges; the cost effective acquisition of multiple datasets for complex engineering components and then the storage, registration, visualisation and analysis of that data. In this respect, there are both experimental and software issues. As illustrated in Figure 2, the main concepts of the body scanner can be separated into two distinct categories

¾ the hardware required for collecting geometrical information, accelerating component location during measurement and thus ensuring the necessary registration of the final data;

¾ the software to establish a generic data repository for data gathering and collation, and to allow geographically distributed co-visualisation.

Figure 3: Combination of different instrument measurement types

Our body scanner data acquisition and visualisation system has been developed to act as a component-centric approach that starts by measuring the important geometrical aspects of and visualisation of different datasets collected from different measurement stations as well as the comparison with datasets containing predictions of 3D (finite element) computer the sample very precisely with a Co-ordinate

Measurement Machine (CMM). At the same time, the physical configuration for different measurement instrument environments (work cells) can be imported into the system to assist with pre-measurement planning. In this way the coordinate axes and extent of the sample are defined and input into the geometry file along models of the same component.

Whilst specific software packages exist to visualize 3D datasets, such as geometric CAD files and X-ray tomograms; these rarely allow the registration and co-visualisation of datasets measured on different types of instruments or the side by side comparison of measurements with finite element (FE) model predictions. In with key fiducial points or features. This geometry file is stored in the central data repository server along with other data files collected subsequently from other instruments.

The registration process is performed semiautomatically with the aid of a registry server to generate an eXtensible Markup Language

(XML)-based master file. This XML-based master file acts as a sample-centric framework that contains multiple datasets along with the

3D geometry. These sets of data follow the sample through the characterisation process.

For any given technique, the registration procedure ensures repeatable registration and alignment of the sample. With the sample geometry stored, the system provides an XML knowledge-based system to assist the materials scientist in the systematic gathering and collation of data. Furthermore, measurement schemes can then be defined in the sample frame, rather than that of the instrument, and similar schemes can be applied on different instruments, or to ensure that identical locations are measured for different sample orientations.

In addition, it acts as a transparent and distributed data management tool for the fusion fact combining the data is usually a major undertaking because different co-ordinate systems and sampling locations have been used for the different experimental conditions. As a result opportunities for bringing together results from different measurement techniques are often missed. This may be because information is stored in different formats, on separate software using different co-ordinate systems. A single XML-based master file contains information about the co-ordinate system and worksheets corresponding to each of the experimental and numerical datasets acquired for the object. This master file removes the registration problem and specialist integration software enables analysis and combination of datasets, for example to calculate stress fields by combining strain components for subsequent visualisation.

In our data environment the registration process is performed semi-automatically with the aid of a registry server to generate the XMLbased master file. This is based on the various experiments generating as a series of data structure types, as illustrated in Figure 3 . The

XML Schema is constructed to allow multiple

High-level of XML schema Snippet of XML master file

<EBS xsi:noNamespaceSchemaLocation="EBS.xsd">

+ <centre_details>

<samples>

<sample id="IDsample1" name="Ti SiC composite" />

</samples>

<datasets>

<data id="vol" name="Tomogram" file="3p6r.vol" />

<data id="strain1" name="Strain map 1" … />

<data id=" strain2" name="Strain map 2" … />

<data id=" strain3" name="Strain map 3" … />

<data id=" strain4" name="Strain map 4" … />

<data id=" strain5" name="Strain map 5" … />

</datasets>

+ <instruments>

<experiments>

+ <tomogram id="vol" … >

<strainmap id="strain1" … >

<data datatype="float32" xdim="29" ydim="2" … />

<transformation>

<Line>

<origin … />

<direction … />

<scale … />

</Line>

</transformation>

</strainmap>

+ <strainmap id="strain2" … >

+ <strainmap id="strain3" … >

+ <strainmap id="strain4" … >

+ <strainmap id="strain5" … >

</experiments>

</EBS>

Figure 4:High-level of XML schema and corresponding example of XML-based snippet of master file hierarchical files nested within each dataset, linked by the simple construction of references.

Existing XML-based master files can be updated with web-based form-filling, or using drag-n-drop features of the XML-editor. The materials scientist can now edit the master file without an extensive knowledge of XML syntax to fuse and manipulate datasets. The master file conforming data files may contain extensive details about the experiment, the sample, geometrical descriptions (with links to external digitised topography), resultant raw data and also post-processed data. Very little data is contained within the XML files themselves, to permit users complete freedom over the choice of analysis packages.

Our XML-based master file contains information about the co-ordinate system and worksheets corresponding to all the experimental and numerical data acquired for the object. This master file removes the registration problem and specialist integration software is written to integrate the different strain field components and to calculate the resulting stress field for subsequent visualisation. Figure 4 has shown the highlevel XML schema and the snippet code of the master file for data relating for a Ti/SiC f composite that contains information about an

3D X-ray tomogram and residual stress data for each fibre, the resulting co-visualisation is shown in scenario A under Figure 5 .

Visualisation of data can be performed using any server that can read the XML-based master file on a local standalone machine.

To date, we have created an XML parser for modular-based

TGS-AMIRA [4] to allow interpretation of the

XML-based master file and rendering of data in

3D or via cross-sections that can be selected using the global sample co-ordinate system. In this way geometric, strain and damage data can be co-visualised (overlaid) or compared with finite element predictions. Additionally, we have taken advantage of SGI OpenGL

VizServer [6] to conduct remote rendering experiments, allowing interactive rendering between geographically distributed material scientists. This means that data being acquired at one location can be visualised and discussed across the network, so that a team can steer the measurement process. Nonetheless, because

TGS-AMIRA does not allow us to take full advantage of a multiple processor system, this has made large and complex dataset analysis

Scenario A: A combination of 3D microstructural information on 140

µ m SiC fibres in a Ti matrix in the form of X-ray tomographic data with fibre elastic strains measured by synchrotron diffraction (FE data can also be available for comparison) [2].

Scenario B: Side by side visualisation of the longitudinal stresses in a railway rail measured by neutrons and the contour method [3]

Figure 5: Two examples on data fusion and side-by-side visualisation and co-visualisation impractical. As part of our on-going development, we have created an

XML parser for customise Visualization Toolkit

(VTK) [5] to act as a visualisation server from

3.

Potential of Engineering Body

Scanner on the Grid

the XML-based master file to perform compressed off-screen rendering and we have visualised the rendered images using a Javabased viewer on the web-viewer in 3D.

The Grid has been characterised by a three-tier model consisting of

¾

Computation/Data grid supports access to the underlying computer system regardless of the location;

¾ Information grid supports content analysis of the data stored or generated using this system;

describes the knowledge about the specific dataset. This is important so as to ensure that it is possible to identify where the data came from and what it represents. The existing XMLbased master file can be used along with a set of

Figure 6: Proposed EBS architecture mounted a Grid Infrastructure

¾ Knowledge grid provides high-level services that can pass the data generation or retrieval from the information grid to fulfil the requirements of the users.

Measurements can be made and analysed globally in different research centres. These data should be collected, merged to existing anticipate a distributed archiving system which includes some form of metadata format that datasets and be available among the participating materials scientists and engineers.

As Figure 6 illustrates a grid-based instruments can benefit from a distributed grid-based data repository, however, these processes and file formats demand a consistent approach, where users share the same standard to extract data metadata to enable a shared open standard among the instruments or techniques for residual stress and damage measurement. A possible enhancement of the XML-based archiving system would be to facilitate the acquisition of knowledge through some simple query answering system against the XML-based from the measurement instrument to be analysed and to make comparison with another techniques. One can also upload the analysed results to make them available for others to compare and vice-versa. Therefore, the information and knowledge must be available either on the centralised location or the distributed repository at all times within the grid environment.

With the current defined XML-based master file format within the EBS system, we can documents in the domain of the residual stress measurements. With this enhanced grid-based repository, materials scientists anywhere could query, search and retrieve the details of a particular component through the grid-based data repository via a globalised automated search engine. When data is made available on the grid, measurements from various techniques and instruments comprise not just sets of results, but also a full description of the data is carried with it as it propagates away from its original

source. The grid-based data repository must ensure that the XML-based master file is gridcompatible, conforming to certain metadata rules and formatting conditions.

The existing EBS concept has the potential to fit into all the above three tiers of grid characterisation, whereby one can

1.

Acquire data from different instruments or techniques (resources-tier) without a need to know about the instrument coordinate system;

2.

Feed the data from different resources into a distributed data repository

(information-tier) via a commonly recognised XML format together with appropriate metadata files;

3.

Analyse and synthesise the data

(knowledge-tier) at geographically separate locations via the distributed visualisation rendering services.

The engineering body scanner development has been targeted to allow the acquisition of datasets from different instruments. This potentially grid-based residual measurement resource can be shared remotely and visualised as a set of “virtual experiments”. The remote

“virtual experiments” will allow the feedback of an XML control file to permit the grid instrumentation of residual stress measurement.

At present, material scientists must meet in a centralised location to analyse, process and discuss the experimental procedures and results.

However, our grid-enabled experiments will allow distributed piping of 3D visualisations to engineers elsewhere, to aid the experimental process and enable active control and contributions from geographically distributed material scientists and engineers.

4.

Future Work on Existing the

Body Scanner System

It is important to note that 3D co-visualisation for the existing EBS system has been aimed at a single, standalone environment although this can be extended to a grid-based remote display environment integrated within an enhanced version of the XML-based data repository. It is vital to capture the feel, frustrations and ambitions of the materials scientist as a user towards the present standalone system before moving on to an enhanced grid-enabled system.

This would involve a structural review to establish measurement criteria. This is because the geographically distributed instruments that we are preparing to use to integrate the present

EBS system into the Access Grid [7], allows interactive discussion within the “virtual experiments”. Discussion is also ongoing with another Grid-based project, Resource-Aware

Visualization Environment (RAVE) [8] allowing collaborative rendering environments for visualisation. Therefore, initially our future work will focus on user evaluation with regard to the usability issues and the extent to which one can interact with the system. This knowledge will then be implemented before extending to a grid-based system.

5.

Acknowledgements

Funding of the EPSRC as well as that provided by a Royal Society-Wolfson Merit Award

(PJW) is gratefully acknowledged.

6.

References

Measurements of Residual Stress in Ring &

Plug, Versailles Project on Advanced

Materials & Structures TWA20, 64 pages

Technology Trend Assessment No. 38, ISSN

1016-2186

[2] S.A. McDonald, M. Preuss, E. Maire, J.-Y.

Buffiere, P.M. Mummery and P.J. Withers,

X-ray tomographic imaging of Ti/SiC composites, Journal of Microscopy, Vol.

209, Pt 2 February 2003, pp. 102–112

[3] J. Kelleher, D. Buttle, J. Shackleton, P.J.

Webster, D.J. Hughes, P.J. Withers, and P.

Mummery, The Effects of Service Lifetime and Wear on the Residual Stress in Railway

Rails, Proceedings of the 6th International

Conference on Engineering Structural

Integrity Assessment, Manchester, 7-9th

October 2002.

[4] TGS-Amira documentation and resources, http://www.tgs.com/support/amira_doc,

2004 http://www.vtk.org/index.php, 2004

Technical White Paper: OpenGL

Vizserver

TM

3.1: Application-Transparent

Remote Interactive Visualization and

Collaboration, 2004

[7] L. Childers, T. Disz, R. Olson, M. E. Papka,

R. Stevens, and T. Udeshi, Access Grid:

Immersive Group-to-Group Collaborative

Visualization, Immersive Projection

Technology, Ames, Iowa, 2000

[8] I.J. Grimstead, N.J. Avis, and D.W. Walker,

RAVE: Resource-Aware Visualization

Environment, All-Hands Meeting, 2004

Download