The research for this thesis is concerned with the improvement of fully automated Multidisciplinary Analysis system. A common approach for testing implementations during run-time of a continuous process is action research [Williamson et al., 2002]. Action research describes one or several loops of analysis of a situation, development of an implementation, implementing it and then evaluating the impact of the implementation. A similar approach is the Design Research Methodology (DRM) as developed by [Blessing and Chakrabarti, 2009]. For this study, DRM is chosen as the methodological research framework, since it extends the scope of action research with a tool to concretely measuring the studies impact. This chapter explains DRM and how it is applied in this research.
« Considering that engineering and design are among the fastest growing fields of sciences, there is surprisingly little methodology to analyse the actual impact and quality of it » [Blessing and Chakrabarti, 2009]. Putting an end to this deplorable state of aﬀairs, [Blessing and Chakrabarti, 2009] developed the Design Research Methodology. The research in this thesis is, with small deviations, structured after their example.
DRM is sectioned in four phases, that can, if necessary, be repeated. Similar to the cyclical engineering approach after [Wieringa, 2005] containing the five optionally repeated phases Problem Analysis, Solution Specification, Solution Analysis, Solution Implementation and Implementation Analysis, the DRM cycle is structured in four phases. The engineering cycle is displayed in Figure 3.2. The four phases of DRM shown in Figure 3.1 are explained in the following chapters.
Since the desired improvement is not necessarily explicitly measurable, the researcher usually has to introduce measurable success criteria. The selection of criteria should be capable of measuring negative as well as positive impacts of the implementation. To properly define the criteria is important for the entire study. As described by [Blessing and Chakrabarti, 2009] they serve to:
identify the aim that the research is expected to fulfil and the focus of the research project;
focus Descriptive Study I on finding the factors that contribute to or prohibit success;
focus the Prescriptive Study on developing support that address those factors that are likely to have most influence;
enable evaluation of the developed support (Descriptive Study II)
Though the researcher has to be aware that their implementation does not necessarily directly cause the change in success criteria. Possible side-eﬀects or other causes have to be evaluated as possible sources for a change in criteria.
The initial overall goal of this study is to find the current challenges and suggest improvements of Multidisciplinary Analysis (MDA) systems by the example of Engineering Workbench at GKN. Reasons for using a MDA is to increase the product development eﬃciency; create more customer value in less time. It is not considered measurable and is therefore represented with other criteria which are identified during the work.
First Descriptive Study
After obtaining an understanding of the problem from literature, an image of the current situation is generated. By combining the results from literature and observation, the factors that influence the success criteria are identified.
The descriptive study can be found in this thesis in the chapter 4.2. The information about the current situation is gathered through semi- and unstruc-tured interviews [Williamson et al., 2002] with employees of GKN relating to the Engineering Workbench
For semi-structured interviews a number of questions are prepared and provide some structure to the interview. Five key employees stretching from developers to users of the Engineering Workbench are interviewed in this fashion; domain experts within thermal analysis (developer), parametric CAD development (developer), design of experiment (user) and computational fluid analysis (developer).
Unstructured interviews are performed as the opportunities appear at the coﬀee machine or while walking in the corridors.
The actual structure of the diﬀerent software modules is recreated by analysing and testing the respective scripts and codes as well as reading the related documentation. Furthermore, literature research is conducted to be able to formally categorise the findings.
The findings are mapped in a standardised modelling language to be able to communicate the results.
Possible faults and/or lack in functionality are noted. The choice of which challenge to address is based on the diﬀerent success criteria and the delimi-tations at present (e.g. time, available software and personal, see Chapter 1.4).
The prescriptive study describes the desired status. One or several methods are developed to improve the status found in the descriptive study. Those methods are then implemented and tested. This is, in a research project, usually done via a demonstrator prototype or proof of concept. A demonstrator contains all implementations necessary to evaluate the impact of the method. A well developed prototype can be used as a basis for future implementation.
This is also the case here; a desired state of the suggested improvements are presented (Figures 4.3 and 4.4) and a prototype developed to evaluate its eﬀectiveness in succeeding to positively aﬀect the associate success criteria.
The prescriptive study for the improved system robustness of EWB is based on literature research for each single problem as identified in the first descriptive study as mentioned in Chapter 4.2.1. Finding the opportunity for improvement is dependant on the nature of each single issue. Among those is the logical deconstruction and rebuilding of the existing code, literature research to already known issues on this topic and producing new program elements. Those code elements were developed and tested for functionality in a separated loop before applying them as part of the prescriptive study.
The method used for the development of a fully automated producibility as-sessment within the Engineering Workbench is to collect information through unstructured interviews. With this information find the functional require-ments of the system and describe a viable process. Finally testing the process by developing and running a prototype system.
Literature research is a continuous process providing valuable insight in prior work within aﬀected fields. Semi- and unstructured interviews are also continuously performed as the need arises. Once every week the progress is evaluated together with the supervisor at GKN.
Second Descriptive study
« Evaluation of design support is a complex, challenging task » [Blessing and Chakrabarti, 2009].
The second descriptive study is in its nature similar to the first descriptive study. It has to be performed with the same rigour and methods.
The goal of the second descriptive study is to evaluate the eﬀect of the implementations of the prescriptive study. Therefore two main questions have to be answered:
Did the implementation of the proposed method into the current working environment succeed?
Did the application/use of the implemented method improve the situa-tion according the the success criteria?
While evaluating the situation, caution is taken with respect to uncontrolled influences. Examples of which are company politics, software updates or other side eﬀects.
The evaluation of EWB after the implementations of methods suggested in the prescriptive study is done as described in the following sections. The second descriptive study of this thesis can be found in Chapter ??.
Most implementations were assessed in their individual impact first, and only later the entire package is evaluated as a whole. The evaluation itself is first done in via test runs with a limited number or only one single design-case. The interaction of the engineers working with EWB is observed, and their feedback is captured in interviews. As the final and most crucial element of testing, one study with 128 design-cases is run on a new TRS design. The study is used in a business case at GKN, and therefore provided suitable feedback.
The final evaluation of the suggested producibility assessment is made by measuring the identified success criteria variables as well as presenting it to aﬀected employees and listening to their feedback.
This chapter combines the description of the original Engineering Workbench (EWB) with the suggestions for improvement. The improvements are mea-sured by the success criteria stated in the first section. Furthermore it explains the implementations of those improvements in detail.
With the implementation of new methods, the performance of EWB should improve in the following points:
Increased Usability The setup of a study shall be simpler and faster. The user should be able to customise the extend of the study to their needs. Users without extensive knowledge in programming or the software packages involved should be able to setup and perform a study. Nonetheless, a certain amount of the involved disciplines will always be necessary to achieve a useful result and interpret it. The goal is to increase the:
Number of potential users
Analysis lead time The time between the setup and the first results. Also the overall time of one whole study should be reduced. The quantity to be limited is the:
Set up time for an entirely new study
Time to set up a re-run of a study with changed parameters
Increased Process Reliability Once a study is started, it should continue until all requested activities have been performed. The failure of single design cases should not jeopardise the entire study. Expressed in a quantifiable way, this results in a reduction of the:
Number of unwanted process terminations
Expanded Functionality Assessment of producibility shall be possible. The number of « loop-backs » between design and production should be decreased. The success is measured by an increase of the:
Number of manufacturing constraints caught. Number of producibility metrics caught.
As an initial step, the original EWB is mapped in UML and represented in use-case and activity diagrams. The use-case diagrams illustrate the setup process of EWB, whereas the activity diagrams describe the process of one EWB study. The entire EWB process is shown in Figure 4.1 as a UML activity diagram.
The baseline geometry is varied with the help of Knowledge Fusion (KF, 1.2) classes. What they do is change NX expressions and attributes according to a Design of Experiment (DoE, 2.3) sampling method. NX expressions are used to control a portion of the geometry such as a specific length. NX attributes represent everything which is not related to the shape of the geometry, such as material specifiers or thicknesses of shell-model faces.
As a part of the geometry variation, the wet-surfaces can be varied to study diﬀerent aerodynamic performances. Wet-surfaces are all surfaces in direct contact to the gas flow. This is done using an Excel Macro document, called « Volvane ». Volvane is a GKN-internal script that generates vane geometries based on boundary conditions. These are directly used for the Computational Fluid Dynamics (CFD) meshing and analysis. For use in the geometry variation process however they need to be translated to the right format and then read into NX with the help of KF classes.
1.2 Industrial environment
1.3 Purpose and research questions
2 Theoretical background
2.1 Knowledge Based Engineering
2.2 Set-Based Concurrent Engineering
2.3 Design of Experiment
2.4 Simulation driven design
2.5 Manufacturability Assessment Systems
3.1 Success criteria
3.2 First Descriptive Study
3.3 Prescriptive Study
3.4 Second Descriptive study
4.1 Success criteria
4.2 Initial state
4.3 Process robustness
4.4 Producibility assessment
5 Discussion and conclusions
5.1 Discussion of method
5.2 Discussion of findings
5.4 Further work
GET THE COMPLETE PROJECT
Multidisciplinary analysis of jet engine components Development of methods and tools for design automatisation in a multidisciplinary context