FPSO Chapter Viii

Embed Size (px)

Citation preview

  • 8/18/2019 FPSO Chapter Viii

    1/36

    CHAPTER VIII.

    CLOSURE 288

    8.1 DEVELOPING THE HPPRM – ADDING NEW KNOWLEDGE?...........................................290

    8.1.1 Revisiting the Research Questions and Hypotheses.......................................... 290

    8.1.2 Validating New Scientific Knowledge – an Overview ......................................292

    8.1.3 Testing the Theoretical Structural Validity (TSV) ...........................................295

    8.1.4 Testing the Empirical Structural Validity (ESV)..............................................298

    8.1.5 Testing the Empirical Performance Validity (EPV).......................................... 301

    8.1.6 Testing the Theoretical Performance Validity (TPV) ....................................... 303

    8.2 DEVELOPING THE HPPRM – ACHIEVEMENTS AND CONTRIBUTIONS........................305

    8.2.1 Contributions to the Field of Engineering Design.............................................305

    8.2.2 Original and Significant: Completing the Criteria for a Ph.D. .......................... 305

    8.3 CRITICAL ANALYSIS – RESEARCH LIMITATIONS..........................................................309

    8.3.1 The HPPRM – Limitations to its Constructs and Framework........................... 309

    8.3.2 The HPPRM – Its Robustness ........................................................................ 311

    8.3.3 The HPPRM – Limitations to the Validation Process ......................................312

    8.3.4 The HPPRM – Limitations to Implementation................................................. 313

    8.4 RECOMMENDATIONS – AVENUES FOR FUTURE WORK ...............................................315

    8.4.1 Industrializing the HPPRM – the Industrial Avenue.........................................315

    8.4.2 Improving the HPPRM Constructs – the Academic Avenue............................. 318

    8.5 CONCLUDING REMARKS....................................................................................................320

  • 8/18/2019 FPSO Chapter Viii

    2/36

    ii

    LIST OF FIGURES

    FIGURE 8-1 HOW THE RESEARCH QUESTIONS AND HYPOTHESES CONNECT TO THE HPPRM

    ................................................................................................................................................... 291FIGURE 8-2 HYPOTHESES TESTING GUIDE: WHERE THE ANSWERS ARE PROVIDED........... 294

    FIGURE 8-3 THE THEORETICAL STRUCTURAL VALIDATION PROCESS – A PICTORIAL

    REPRESENTATION .................................................................................................................. 297

    FIGURE 8-4 THE EMPIRICAL STRUCTURAL VALIDATION PROCESS– A PICTORIAL

    REPRESENTATION .................................................................................................................. 300

    FIGURE 8-5 THE EMPIRICAL PERFORMANCE VALIDATION PROCESS – A PICTORIAL

    REPRESENTATION .................................................................................................................. 302

    FIGURE 8-6 THE THEORETICAL PERFORMANCE VALIDATION PROCESS – A PICTORIAL

    REPRESENTATION .................................................................................................................. 304

    FIGURE 8-7 THE CONTRIBUTIONS TO THE FIELD OF ENGINEERING DESIGN .............. ......... 308

    FIGURE 8-8 THE RESEARCH AVENUES STEMMING FROM THE HPPRM FRAMEWORK......... 317

    LIST OF TABLES

  • 8/18/2019 FPSO Chapter Viii

    3/36

    CHAPTER VIII

    8. CLOSURE

    NumTax TechDiff c-DSP HPPRM Closure, demonstrating

    • research validity

    • research originality

    • research significance

    XXX Xx

    x x

    x x

    x x

    x x

    x x

    x

    In this dissertation a framework has been developed, presented, and tested to facilitate the

    realization of large, complex and expensive MTO systems that are traditionally produced in small

    numbers using much of the same technology. Our objective in this chapter is then to bring the

    development and presentation of this framework to a closure. By closure we mean to demonstrate that we

    have achieved the principal research objective and answered the questions we posed in Chapter I, in a

    satisfactory manner.

    We do this by first demonstrating (by means of the validation square) that we have added newknowledge to the field of engineering design. Then, we evaluate this new knowledge in terms of its

    originality and significance, in order to fully demonstrate that this research comply with the requirements

     posed for the Doctor of Philosophy degree in Mechanical Engineering.

    Finally, we discuss the limitations of the HPPRM framework in terms of its constructs, its

    robustness, its validity, and its industrial implementation. Based on this, we give recommendations on

    “This must n ot be taken as the end; 

     i t may poss ib ly be the beginn ing of the end; 

     but i t is cer ta in ly the end of the beginn ing”  – Winston Churchill

    In speech at the Mansion House in London City in late 1942

  • 8/18/2019 FPSO Chapter Viii

    4/36

    CHAPTER VIII

    Closure

     

    289

    future research building on this framework, both in academia and in industry, before we close this

    dissertation.

  • 8/18/2019 FPSO Chapter Viii

    5/36

    CHAPTER VIII

    Closure

     

    290

    8.1 DEVELOPING THE HPPRM –

    ADDING NEW KNOWLEDGE?

    In this section we demonstrate by means of the validation square (see Section 1.6.2) that the

    HPPRM gives reliable and useful output for valid input, hence, that it adds new knowledge to the field of 

    engineering design. In Section 8.1.1 we revisit the research questions and hypotheses that were posed in

    Section 1.5. For reference, in Figure 8-1 we show where the research questions and hypotheses are

    embedded in the HPPRM structure. In Section 8.1.2 we give an overview of the procedure for 

    substantiating that new scientific knowledge has been added to the field of Engineering Design, and in

    Sections 8.1.3 through 8.1.6 we demonstrate that that new scientific knowledge has been added to the

    field of Engineering Design. Adding originality and significance in Section 8.2, we can assert compliance

    with the Ph.D. requirements as given by the GWW School of Mechanical Engineering.

    8.1.1 Revisiting the Research Questions and Hypotheses

    Validation in context of Ph.D. research rests on (1) having answered the posed questions, (2)

    the answers being in accordance with the hypotheses, and (3) the answers being acceptable from a Ph.D.

    requirement perspective. Hence, we start the validation procedure presentation by revisiting the questions

    and the hypothesis posed for this research.

    Fundamental Research Question: How can large and complex made-to-order systems produced in smallnumbers using much of the same technology be realized more effectively and efficiently for different

    applications?

    Fundamental Hypothesis: Large and complex made-to-order systems as characterized above can be

    realized more effectively and efficiently for different applications by developing Hierarchical Product

    Platforms (see Section 1.3.3) from an evolutionary perspective (see Chapter II).

    As stated in Section 1.5, the fundamental question has been partitioned into three questions,

    where each question deals with some fundamental aspects regarding how to develop a Hierarchical

    Product Platform from an Evolutionary Perspective. These questions and their corresponding hypotheses

    are formulated as follows.

    Research Question 1: Based on a firm’s existing designs; how can HPPs be defined in terms of what

     physical entities to standardize for which products?

    Hypothesis 1: Numerical Taxonomy (Sneath and Sokal 1973) provides a good framework for defining

    HPPs in terms of what physical entities to standardize for which products (answering Research

    Question 1 acknowledging Path-Dependence).

  • 8/18/2019 FPSO Chapter Viii

    6/36

    CHAPTER VIII

    Closure

     

    291

    … By designing Hierarchical Product Platforms

    from an Evolutionary Perspective, characterized

    by Path-dependence, Population-thinking, and

    Probabilistics.

    … By designing Hierarchical Product Platforms

    from an Evolutionary Perspective, characterized

    by Path-dependence, Population-thinking, and

    Probabilistics.

    How can large and complex made-to-order systems produced in

    small numbers using much of the same technology be realized

    more effectively and efficiently for different applications?

    How can large and complex made-to-order systems produced in

    small numbers using much of the same technology be realized

    more effectively and efficiently for different applications?

    The HPPRM

    Gather Data

    Cluster Data

    For each Assembly Level (AL)

       P   h  a  s  e

       I  :   D   E   F   I   N   E

       (   R  e  s  e  a  r  c

       h   Q  u  e  s   t   i  o  n   1   )

    Establish

    Design Requirements

    Partition Realization

    Processes

       P   h  a  s  e   I   I  :   M   O   D   E   L

       (   R  e  s  e  a  r  c   h   Q  u  e  s   t   i  o  n   2   )

    Model Relationships

    and formulate c-DSPs

    Establish ScenariosSolve for each

    Scenario

       P   h  a  s  e   I   I   I  :   S   O   L   V   E

       (   R  e  s  e  a  r  c   h   Q  u  e  s   t   i  o  n   3   )

    Decide on HPP

    For each Assembly Level (AL)

    For ALL Assembly Levels (AL)

    Hypothesis 1

    Hypothesis 2

    Hypothesis 3

    Numerical Taxonomy

    (Sneath & Sokal 1973)

    Technology Diffusion

    (Silverberg, Dosi et al. 1988)

    Compromise DSP

    (Mistree, Hughes et al. 1993)

    The Research Hypotheses

    THE FUNDAMENTAL QUESTION THE FUNDAMENTAL HYPOTHESIS

    Figure 8-1

    How the Research Questions and Hypotheses Connect to the HPPRM

  • 8/18/2019 FPSO Chapter Viii

    7/36

    CHAPTER VIII

    Closure

     

    292

    Research Question 2: In order to better adapt products to their requirements we want to consider 

    realizing them with alternative processes; how can we evaluate the effects such a transition may have

    on a product’s operational performance, its cost and its schedule, accounting for the learning that has

    to take place in a transition phase?

    Hypothesis 2: Technology Diffusion (Silverberg, Dosi et al. 1988) is a viable concept to include learning

    when evaluating the effects of transferring to new processes (answering Research Question 2 from a

    Population-Thinking perspective).

    Research Question 3: How can products and processes be synthesized in a mathematical model,

    acknowledging that it is based on limited information and ‘volatile’ assumptions?

    Hypothesis 3: The c-DSP (Mistree, Hughes et al. 1993) provides a structure and an anchoring that is well

    suited for solving models that are acknowledged to be based on incomplete information and ‘volatile’

    assumptions (answering Research Question 3 acknowledging the Probabilistic nature of the problem)

    As stated, the validity of the research in context of Ph.D. requirements rests on having

    answered the questions according to the hypotheses in an acceptable way. In this research, the answers

    correspond to the hypotheses, and hypotheses tested valid according to the process given in Section 1.6,

    are asserted to constitute new scientific knowledge. In addition, hypotheses found to be significant and

    original in addition to being valid are considered to be contributions to the field of Engineering Design,

    see Section 8.2.

    8.1.2 Validating New Scientific Knowledge – an Overview

    In general a research question is answered when the corresponding hypothesis is validated.

    However, in this dissertation there is a hierarchical organization of questions and hypotheses, where a

    fundamental question and hypothesis is partitioned into three supporting questions and hypotheses. The

    hierarchical organization implies that the answer to the fundamental question lies in validating the

    fundamental hypothesis; validating the fundamental hypothesis lies in answering the supporting

    questions; answering the supporting questions lies in validating the supporting hypotheses. Hence, in

    order to answer the fundamental question, we validate each of the supporting hypotheses. This is done

    according to the process given in Section 1.6, where each of the quadrants in the

    Validation Square to the right are tested separately for each supporting

    hypothesis. Then, by adding additional information, the validity of the

    fundamental hypothesis – i.e., the validity of the HPPRM – is asserted through

    induction. In the following, we present the procedure for testing each hypothesis

    according to the process outlined in Section 1.6, and in Figure 8-2 we present a

    THEORETICAL

    STRUCTURAL

    VALIDITY

    EMPIRICAL

    STRUCTURAL

    VALIDITY

    EMPIRICAL

    PERFORMANCE

    VALIDITY

    THEORETICAL

    PERFORMANCE

    VALIDITY

  • 8/18/2019 FPSO Chapter Viii

    8/36

    CHAPTER VIII

    Closure

     

    293

    guide to where the various testing takes place in the dissertation (where the answers to the questions are

     provided).

  • 8/18/2019 FPSO Chapter Viii

    9/36

    CHAPTER VIII

    Closure

     

    294

    Ch. I

    NumTax TechDiff c-DSP HPPRM Demonstrating• Originality of the HPP: Sc. 1.3.5

    Ch. V

    NumTax TechDiff c-DSP HPPRM

    x

    Ch. IV

    NumTax TechDiff c-DSP HPPRM

    xXx x

    Xx x

    Xx x x

    Demonstrating that constructs are

    • accepted in general: Sc. 4.1.5, 4.2.4, 4.3.3

    • accepted in HPPRM: Sc. 4.1.6, 4.2.5, 4.3.4

    • useful in HPPRM: Sc. 4.1.7, 4.2.6, 4.3.5

    X X X x

    Ch. VI

    NumTax TechDiff c-DSP HPPRM

    X X x x

    Case Study, demonstrating that

    • HPPRM fits problem: Sc. 6.1• usefulness of Num Tax: Sc. 6.6

    Ch. VIII

    NumTax TechDiff c-DSP HPPRM Closure, demonstrating

    • research validity: Sc. 8.1

    • research originality: Sc. 8.2.2

    • research significance: Sc. 8.2.2

    XXX Xx

    x x

    x x

    x x

    x x

    x x

    x

    Ch. III

    NumTax TechDiff c-DSP HPPRM Demonstrating internal consistency

    of the HPPRM by Flow Charts

    - Phase I Define: Sc. 3.2

    - Phase II Model: Sc. 3.3- Phase III Solve: Sc. 3.4

    x

    Ch. VII

    NumTax TechDiff c-DSP HPPRM

    X X x

    Case Study, demonstrating

    • usefulness of Tech Diff: Sc. 7.4.5

    • usefulness of c-DSP: Sc. 7.5

    • usefulness of HPPRM: Sc. 7.10

    NumTax TechDiff c-DSP HPPRM

    Ch. II

    Demonstrating applicability of

    • Evolutionary approach: Sc. 2.1-3

    • Numerical Taxonomy: Sc. 2.4

    • Technology Diffusion: Sc. 2.5

    • c-DSP: Sc. 2.6X X X x

    xxx xxx

    x

    Demonstrating internal consistency

    of the HPPRM by Example Problem

    - Phase I Define: Sc. 5.2

    - Phase II Model: Sc. 5.3

    - Phase III Solve: Sc. 5.4x x xx x x

    x

    x

    Figure 8-2

    Hypotheses Testing Guide: Where the Answers are Provided

  • 8/18/2019 FPSO Chapter Viii

    10/36

    CHAPTER VIII

    Closure

     

    295

    8.1.3 Testing the Theoretical Structural Validity (TSV)

    The hypothesis-testing starts with the Theoretical Structural

    Validity (TSV) – the main pillar in the bridge over the ‘chasm of new

    knowledge’ – and the process of demonstrating the TSV of the HPPRM is

    illustrated in Figure 8-3.

    In context of a design method, this validity rests on (1) the

    ‘correctness’ of the individual constructs constituting the complete method and (2) the internal

    consistency in the way the constructs are organized within the method. Together this validates that the

    results are obtained in a correct and consistent manner, implying that for valid input the output can be

    trusted. The specific approach to demonstrate ‘correctness’ of the individual constructs is by literature

    referrals, and the specific approach to demonstrate internal consistency is by flow-chart representation of the method.

    The design method to be tested is the HPPRM. Embedded in this method are the three

    hypothesized constructs, namely, Numerical Taxonomy, Technology Diffusion, and the compromise DSP.

    Hence, testing the Theoretical Structural Validity of the HPPRM lies in testing the Theoretical Structural

    Validity of each of these constructs in addition to testing the internal consistency of the HPPRM. The

    Theoretical Structural Validity of the key constructs is demonstrated as follows.

    §  The Theoretical Structural Validity of Numerical Taxonomy   is demonstrated in Section 4.1 in

    general and in Section 4.1.5 in particular. This is done by demonstrating that Numerical Taxonomy not only is generally accepted for revealing the inherent structures / pattern in data-sets, but it is used

    to benchmark new methods as well.

    §  The Theoretical Structural Validity of Technology Diffusion   is demonstrated in Section 4.2 in

    general and in Section 4.2.2 in particular. This is done by demonstrating that Technology Di ff usion 

    is accepted for simulating the impact on market shares when firms transfer to new technology within

    the same industry.

    §  The Theoretical Structural Validity of the compromise-DSP  is demonstrated in Section 4.3 in general

    and in Section 4.3.2 in particular. This is done by demonstrating that the compromise-DSP   is

    generally accepted as a construct for multi-objective optimization used in searching for ‘satisficing’

    rather than optimal solutions.

    The internal consistency of the HPPRM is demonstrated in Sections 3.2 through 3.4, where

    each phase is elaborated and visualized in a separate flow-chart. In doing so, the information flow is

    emphasized and the consistency is demonstrated in terms of (1) there is no redundant information being

    generated and (2) the underlying assumptions are identified, evaluated and found valid.

          T      S      V

    EPVTPV

          E      S      V

  • 8/18/2019 FPSO Chapter Viii

    11/36

    CHAPTER VIII

    Closure

     

    296

    Taking the Theoretical Structural Validity of the hypothesized constructs together with the

    internal consistency of the HPPRM, it is asserted that the HPPRM is Theoretical Structural Valid.

  • 8/18/2019 FPSO Chapter Viii

    12/36

    CHAPTER VIII

    Closure

     

    297

    Numerical

    Taxonomy: 

    Sneath & Sokal, 1973

    Keller, 1992

    …ü

    Technology 

    Diffusion: 

    Hall, 1994

    Silverber, Dosi et al. 0988

    …ü

    c-DSP; Mistree et al. 1990

    Lewis et al. 1995

    Chen et al. 1996

    …ü

    HPPRM 

    Theoretical

    Structural

    Validü

      CRITERIA

    Theoretical Structural Validity 

      Tools / Methods Used in

    HPPRM Considered Valid

      within their Specified Ranges,  i.e., for Certain Applications

      The Internal Structure of

    HPPRM Considered

    Consistent

     ü

     ü

    I

    I

    I

    TI

    I

    I T

    I

    TI

    I T

    ?

    T

    I

    I

    I

    I

    T

    TI

    I

    T

    I

    T

    I

    I I

    I T

    I

    I

    I

    T

    T

    I

    I

    T

    I

    T

    I

    I

    I

    Figure 8-3

    The Theoretical Structural Validation Process – 

    A Pictorial Representation

  • 8/18/2019 FPSO Chapter Viii

    13/36

    CHAPTER VIII

    Closure

     

    298

    8.1.4 Testing the Empirical Structural Validity (ESV)

    The hypothesis-testing continues with the Empirical Structural

    Validity (ESV) – the other pillar in the bridge over the ‘chasm of new

    knowledge’ – and the process of demonstrating the ESV of the HPPRM is

    illustrated in Figure 8-4.

    In context of a design method, this validity refers to (1) applying

    the method within acceptable ranges of the method-constructs, (2) the proposed case studies being

    appropriate, and (3) adequate quality of the data associated with the case studies. Together this validates

    that the problem fits the method, that the case studies fit the problem, and that the testing fits the case

    studies, i.e., that the input intended for the method is valid. The specific approach to demonstrate that the

     problem fits the method is by comparing the intended problem to the problems for which the constructsare generally accepted. The specific approach to demonstrate that the case studies fits the problem is by

    comparing the characteristics of the chosen case studies to the characteristics of the general problem as

    given in Section 1.5.1. The specific approach to demonstrate that the available data for the case studies

    have sufficient quality to support a conclusion is by induction.

    First of all, does the problem fit the method?  In other words, is the problem suited for the

    constructs constituting the method? This is elaborated in Chapter IV, where the applicability of each of 

    the key constructs in the HPPRM is critically evaluated. The applicability of Numerical Taxonomy   is

    elaborated in Section 4.1.6; the applicability of Technology Di ff usion  is being elaborated in Section 4.2.5;

    and the applicability of the c-DSP  is being elaborated in Section 4.3.4.

    Secondly and thirdly, do the case studies fit the problem, and does the testing fit the case

     studies? In other words, are the case studies appropriate for demonstrating the usefulness of the method?

    For the purpose of testing the method performance, one Example Problem and one Case Study is used,

    namely, a Family of Gravitational Separators (Chapter V) and a Family of Marginal Field Vessels

    (Chapter VI and VII). The appropriateness of the Example Problem is elaborated in Section 5.1.3,

    demonstrating that the Gravitational Separators are representative of the general problem and that the

    data, being ‘tailored’ to test the clustering algorithm, is suited to demonstrate the methodology. Note that

    even if the data is ‘tailored’ to test the clustering algorithm, it also demonstrates the usefulness of the

    method for the hypothetical design portfolio. The appropriateness of the Case Study is elaborated in

    Section 6.1.2, demonstrating that the Marginal Field Vessels are representative of the problem and that

    the data, being real life data, is suited to demonstrate the usefulness of the method.

    EPVTPV

          E      S      V

          T      S      V

  • 8/18/2019 FPSO Chapter Viii

    14/36

    CHAPTER VIII

    Closure

     

    299

    Given that the Gravitational Separator example problem and the Marginal Field Vessel case

    study are appropriate for testing the Empirical Performance Validity of the HPPRM, we assert that the

    HPPRM is valid for its intended input, hence, the HPPRM is Empirical Structural Valid.

     ü

      CRITERIA

    Empiri cal Structural Validity 

      Intended Application ofTools / Methods in HPPRM

      within their Specified Ranges

     

    Examples / Case studies

    Representative for Problem

      Able to Support Conclusions

     ü

     ü

    Numerical

    Taxonomy: Sneath & Sokal, 1973

    Keller, 1992

    Valid Applications

    Classification

    Pattern recognition

    … ü

     ü

    Technology 

    Diffusion: Hall, 1994

    Silverber, Dosi et al. 0988

    … 

    Valid Applications

    Discounting Performance

    Learning

    … ü

     ü

    c-DSP; Mistree et al. 1990

    Lewis et al. 1995

    Chen et al. 1996

    Valid Applications

    Multi-Objective optimization

    Distance to Target

    … üü

     ü HPPRM TheoreticalStructural

    Valid

    Empirical

    Structural

    Validü

     ü

     ü

  • 8/18/2019 FPSO Chapter Viii

    15/36

    CHAPTER VIII

    Closure

     

    300

    Figure 8-4

    The Empirical Structural Validation Process– 

    A Pictorial Representation

  • 8/18/2019 FPSO Chapter Viii

    16/36

    CHAPTER VIII

    Closure

     

    301

    8.1.5 Testing the Empirical Performance Validity (EPV)

    Having demonstrated that the HPPRM is likely to provide

    trustworthy output for its intended and validated input, the next step is to

    validate the usefulness of this output; i.e., to test the Empirical Performance

    Validity (EPV). This is viewed as the first stone paving the bridge over the

    ‘chasm of new knowledge’, and the process of demonstrating the EPV of 

    the HPPRM is illustrated in Figure 8-5.

    In context of testing the HPPRM, EPV refers to whether the HPPRM is able to produce useful

    results for the chosen Example Problem and the chosen Case Study, and whether the key HPPRM

    constructs (Numerical Taxonomy, Technology Diffusion and the compromise-DSP) contribute positively

    to this usefulness. Usefulness in this context refers to whether the resulting HPP contributes to reducingcost and time while maintaining acceptable operational performance (increase competitiveness).

    First of all, the usefulness of the Hierarchical Product Platform (HPP) synthesized from the

    HPPRM is demonstrated in Section 5.4.4 for the Example Problem and in Section 7.10 for the Case Study.

    Having answered the major performance validity question, the remaining issue is whether the key

    constructs are contributing to this usefulness.

     Numerical Taxonomy is anticipated to contribute to the HPPRM usefulness by identifying the

     potential for standardization in an existing design portfolio, hence, reducing the combinatorial problem to

    a manageable size. This is demonstrated in Section 4.1.7.

    Technology Diffusion is anticipated to contribute to the HPPRM usefulness by facilitating

    evaluation of alternative processes for various steps in the fabrication process, hence, advocating a

    continuous improvement approach. This is demonstrated in Section 4.2.6.

    The compromise DSP is anticipated to contribute to the HPPRM usefulness by integrating the

     product and process model, hence, facilitating solving for multiple objectives and advocating robust and

    adaptive solutions. This is demonstrated in Section 4.3.5

    Having demonstrated that Hierarchical Product Platforms are useful in terms of saving time

    and cost for large and complex made-to-order systems, and that the key constructs in the HRRPM

    contributes to this usefulness, we assert that the HPPRM is useful at least for the chosen Example Problem

    and Case Study. Hence, we assert that the HPPRM is Empirical Performance Valid,

    EPV

          T      S      V

    TPV

          E      S      V

  • 8/18/2019 FPSO Chapter Viii

    17/36

    CHAPTER VIII

    Closure

     

    302

      c-DSP: 

      “Probabilistic”

    Purpose:

      Gain appreciation .

      about problem to

      support decisionüü

     ü

      Technology 

      Diffusion: 

    Hall, 1994  “Population-

    …  Thinking”

      Purpose:

      Discount process

      performance as

    learning takes

     place

     ü

     ü

     ü

      Numerical

    Taxonomy: 

    Sneath, 1973

    Keller, 1992 “Path-… Dependence”

    Valid App…  Purpose:

    Classification   Find potential for 

    Pattern rec…  standardization

    …  within existing

     portfolio ü

     ü

      CRITERIA

    Empir ical Perf ormance Vali dity 

      Numerical Taxonomy to find

    potential for Standardization

    Technology Diffusion useful to

    discount process performance

      c-DSP useful to gain apprecia-

    tion about problem

      HPP useful to reduce cost and

    time while maintaining oper. perf.

     ü

     ü

     ü

     ü ü

     ü

    HPPRM 

    Theoretical

    Structural

    Valid

    Empirical   Empirical

    Structural   Performance

    Valid   Validü

     ü

     ü

    ...

     ü

    Figure 8-5

    The Empirical Performance Validation Process – 

    A Pictorial Representation

  • 8/18/2019 FPSO Chapter Viii

    18/36

    CHAPTER VIII

    Closure

     

    303

    8.1.6 Testing the Theoretical Performance Validity (TPV)

    Having demonstrated validity for a selected set of case studies,

    what is left is to build confidence in the validity beyond this set of case

    studies, i.e., to build confidence in generality. This validity is termed

    Theoretical Performance Validity (TPV), and is viewed as the last stone

     paving the bridge over the ‘chasm of new knowledge’. This process of 

    demonstrating the TPV of the HPPRM is illustrated in Figure 8-6.

    In context of testing the HPPRM it refers to the HPPRM being capable of producing useful

    results beyond the Gravitational Separator Family example problem and the Marginal Field Vessel Family

    case study. In order to substantiate a generality claim for the HPPRM, we apply a piecewise strategy: The

    Theoretical Performance Validity of the HPPRM rests on three pillars, namely;

    1.  that the key method-constructs are applicable for problems beyond the Example Problem and

    the Case Study. Their generality and applicability is demonstrated in Sections 4.1.5, 4.2.4 and

    4.3.3, and in Sections 8.1.3 and 8.1.4 as part of the Theoretical and Empirical Structural

    Validation.

    2.  that the Example Problem and Case Study are representative for the general problem. This is

     part of the Empirical Structural Validation, see Sections 8.1.4.

    3.  that the HPPRM is useful for at least some selected case studies. This is part of the Empirical

    Performance Validation, see Sections 8.1.5.

    However, the final acceptance of the HPPRM being Theoretical Performance Valid requires a ‘ leap of 

    faith’ , where the three other bridge elements (Theoretical Structural Validity, Empirical Structural

    Validity, and Empirical Performance Validity) are viewed as ‘leap facilitators’. This is illustrated in

    Figure 8-6, and it concludes the validation of the Hypotheses and hence, the answering of the questions in

    compliance with the requirements of the GWW School of Mechanical Engineering for Ph.D. research.

    If we have been successful in facilitating a ‘leap of fai th ’, this research is accepted as having

    added new scientific knowledge to the field of Engineering Design. Further, we assert that we have

    achieved the principal objective of this dissertation, namely, that we have

    developed a conceptual framework for realizing Hierarchical Product Platforms from which a family

    of products using much of the same technology can be designed and realized for different 

    applications.

    What is left to demonstrate is that the HPPRM is significant and original. This demonstration

    starts with identifying the specific contributions to the field of Engineering Design from developing the

    EPVTPV

    ‘Leap of Faith’

          E      S      V

          T      S      V

  • 8/18/2019 FPSO Chapter Viii

    19/36

    CHAPTER VIII

    Closure

     

    304

    HPPRM, i.e., from achieving the principal objective. Then, we evaluate these contributions in terms of 

    their significance and originality. This is done in the next section.

      HPPRM 

    Theoretical

    Structural

    Valid

    Empirical   Empirical

    Structural   Performance

    Valid   Valid for 

      case study

     ü

     ü

     ü

      CRITERIA

    Empir ical PerformanceValidity 

      HPPRM valid beyond

    example problem and

    case study

     ü

    HPPRM 

    Theoretical Theoretical

    Structural Performance

    Valid Valid

    Empirical   Empirical

    Structural Performance

    Valid Validü

     ü

     ü

     ü

      HPPRM 

    Theoretical

    Structural

    Valid

    Empirical   Empirical

    Structural   Performance

    Valid   Valid for

    example problem

     ü

     ü

     ü

    Through the process we

    hope to have built

    confidence in generality

    “a Leap of Faith ”

    Figure 8-6

    The Theoretical Performance Validation Process – 

    A Pictorial Representation

  • 8/18/2019 FPSO Chapter Viii

    20/36

    CHAPTER VIII

    Closure

     

    305

    8.2 DEVELOPING THE HPPRM –

    ACHIEVEMENTS AND CONTRIBUTIONS

    Assuming that developing the HPPRM adds new scientific knowledge to the field of 

    Engineering Design, what is left to demonstrate (according to the Ph.D. research requirements) is that this

    added scientific knowledge is significant and original. This demonstration is based on (1) identifying

     potential contributions to the field of engineering design, and evaluating these potential contributions in

    terms of their (2) significance and (3) originality. If the potential contributions are found to be both

    significant and original, we assert that developing the HPPRM is a contribution to the field of Engineering

    Design, and hence, this research complies with the Ph.D. requirements as given by the GWW School of 

    Mechanical Engineering.

    8.2.1 Contributions to the Field of Engineering Design

    First of all, in Section 8.1 we demonstrate that we have achieved our principal goal, namely, to

    develop a framework for realizing Hierarchical Product Platforms from an Evolutionary Perspective.

    From this achievement we derive the following contributions to the field of engineering design, see Figure

    8-7:

    1)  the Hierarchical Product Platform Realization Method (HPPRM);

    2)  an Evolutionary Approach to engineering systems development;

    3)  the Validation Square anchored in the relativistic school of epistemology;

    8.2.2 Original and Significant: Completing the Criteria for a Ph.D.

    The three contributions are apparently interconnected, however, we will evaluate them

    separately for their significance and originality. This is based on the assumption that even if one or more

    of the three potential contributions are found not to be significant and / or original, the others still may be.

    The significance and originality are demonstrated in the following.

    The HPPRM – Original and Significant?

    The originality  of the HPPRM lies in the Hierarchical Product Platform concept and the

    method-constructs, namely, Numerical Taxonomy and Technology Diffusion. In Section 1.3.4 we

    demonstrate the originality of the HPP by pointing to how the HPP is an advancement to current product

  • 8/18/2019 FPSO Chapter Viii

    21/36

    CHAPTER VIII

    Closure

     

    306

     platform and product family design. In Section 2.4.3 we demonstrate the originality of Numerical

    Taxonomy in the context of engineering design, by pointing to how it integrates Group Technology,

    clustering and Product Platform thinking. In Section 2.5.2 we demonstrate the originality of Technology

    Diffusion by pointing to its origin being in the economics, and that it have not been applied in engineering

    design before.

    The significance  of the HPPRM lies in the usefulness of the HPP from the perspective of 

    reducing cost and schedule while maintaining operational performance at an acceptable level. In Table 7-

    22 we show the CAPEX savings to be around 6 mill USD and the time saving to be around 8 months for 

    the CAPEX driven HPP. In Table 7-23 we show the CAPEX savings to be around 3 mill USD and the

    time saving to be around 1.5 years for the Marginal Field driven HPP. In context of developing marginal

    fields, we consider these figures to be significant, and hence, we assert that the HPPRM is both an original

    and significant contribution to the field of engineering design.

    The Evolutionary Approach – Original and Significant?

    The originality of the evolutionary approach lies in the fact that it is adopted from an entirely

    different domain, namely, from biology. Even if taken from a different domain, it does not seem to

    contradict conventional knowledge in the field of engineering design, and in Section 2.3.2 we explicitly

    demonstrate that the evolutionary approach actually compliments it.

    Since the HPPRM is based on the evolutionary approach, its significance is directly linked to

    the significance of the HPPRM. The contribution from the evolutionary approach comes from introducing

    a new perspective where adapting to change becomes the imperative objective, and where variation is seen

    as the key to become increasingly adaptive. Based on this, we assert the evolutionary approach to be both

    an original and significant contribution to the field of engineering design.

    TheValidation Square – Original and Significant?

    The originality of the Validation Square lies in the application rather than the origin. The

    Validation Square comes from the realm of system dynamics testing (Richardson and Pugh 1988), then it

    was used in (Bailey 1997), before we expand it in this dissertation to comprise design method testing.

    The significance of the Validation Square lies in its comprehensiveness where all aspects of 

    verification / validation is covered. Based on this we assert that the Validation Square constitutes an

    original and significant contribution to the field of engineering design.

    We have now demonstrated that the HPPRM represents new knowledge to the field of 

    Engineering Design, and that the contributions stemming from developing this framework are original

  • 8/18/2019 FPSO Chapter Viii

    22/36

    CHAPTER VIII

    Closure

     

    307

    and significant. Hence, we assert that this research complies with the Ph.D. requirements as given by the

    GWW School of Mechanical Engineering.

  • 8/18/2019 FPSO Chapter Viii

    23/36

    CHAPTER VIII

    Closure

     

    308

    FOUNDATION FOR RESEARCHFOUNDATION FOR RESEARCH

    THE “EVOLUTIONARY APPROACH”THE “EVOLUTIONARY APPROACH”

    THE “HPPRM”THE “HPPRM”

    BIOLOGY

    Evolution

    ECONOMY

    Technologydiffusion

    ENGINEERING

    Optimization

    NumericalTaxonomy

    (Hypothesis 1)

    EvolutionaryTechnology Diffusion

    (Hypothesis 2)

    Compromise DecisionSupport Problems

    (Hypothesis 3)

     AssemblyTaxonomy

    Performancediscounting

    C-DSP formulationfor HPPs

    HPP definition(RQ1)

    Process Modeling

    incl. learning (RQ2)

    Product

    Model

    HPP c-DSP;an Integrated Product

    and Process Model

    Process

    Model

    Synthesis of Product

    and Process (RQ3)

    Standardization

    Taxonomy

    StandardizationTaxonomy

    Theoretical TheoreticalStructural Performance

    Validity Validity

    Empirical Empirical

    Structural PerformanceValidity Validity

    Theoretical TheoreticalStructural PerformanceValidity Validity

    Empirical EmpiricalStructural PerformanceValidity Validity

    THE “VALIDATION SQUARE”

    PHILOSOPHY

    Epistemology

    Social / holistic /relativistic School

    of Epistemology

    Process of

    building confidencein usefulness

    Figure 8-7

    The Contributions to the Field of Engineering Design

  • 8/18/2019 FPSO Chapter Viii

    24/36

    CHAPTER VIII

    Closure

     

    309

    8.3 CRITICAL ANALYSIS –

    RESEARCH LIMITATIONS

    Our objective in this section is to identify the limitations to the presented research. We do this

     by first evaluating (in Section 8.3.1) the limitations of the method-constructs themselves and what

    limitations this imposes on the total framework. Then, in Section 8.3.2 we evaluate the robustness of the

    framework by asking ‘what-if’ questions regarding the objective (the what) and the approach (the how).

    Then, in Section 8.3.3 we take a critical look at the HPPRM validity to identify its most significant

    weaknesses. Finally, in Section 8.3.4 we evaluate what it will take to implement such a structure in, for 

    example, Kvaerner. The information from this exercise is then used to identify avenues for future work,

    which is dealt with next in Section 8.4.

    8.3.1 The HPPRM – Limitations to its Constructs and Framework

    The HPPRM is a framework intended for identifying a direction into the future based on

    leveraging the past – a process that rests on the following core assumptions.

    §  The HPPRM is developed for the Offshore Oil and Gas industry, based on the existence of a

     portfolio of similarly but not identically configured products;

    §  Standardization improves cost and schedule while compromising operational performance, and

    we are seeking the standardization level that gives the best compromise;

    §  When seeking the standardization level that gives the best compromise, we prefer robust

    solutions to optimized solutions, i.e., we seek solutions that are significantly better than what

    we started out with and not likely to change for relatively small changes in the environment;

    §  In addition to seeking a standardization level, we seek a corresponding process that enhances

    the benefits of standardization, hence, we assume the existence of information to support the

    estimates of learning rates and initial skill levels.

    Based on these assumptions, we analyze the product portfolio, at various levels of assembly, to

    identify where there is a potential for standardization. In other words, we identify parts / assemblies that

    are so similar that it seems justifiable to reduce the number of options to one, i.e., to standardize for this

     particular part / assembly. We recognize that the mentioned similarities can appear at different levels of 

    assembly and can involve different end products at the various assembly levels. Hence, at different levels

    of assembly there can be items standardized for different end products. This is the premise for the

    Hierarchical Product Platform, and the following discussion will be in context of the above. Note,

    however, that this approach is very much linked to how much is invested in the past, hence, the HPPRM

  • 8/18/2019 FPSO Chapter Viii

    25/36

    CHAPTER VIII

    Closure

     

    310

    is not intended for industries where product changes can be made without substantial consequences to the

     project execution.

    Numerical Taxonomy – Limitations

    We use Numerical Taxonomy to identify the potential for standardization by quantifying the

    similarities between the parts we compare. Numerical Taxonomy is a stand-alone method based on one

    initial phase for determining what is to be compared (the OTUs), one intermediate phase for clustering,

    and a final phase for construction of taxa. The first and last phase involves the use of heuristics, which

    have not been formalized in this research, hence, affecting the repeatability. This constitutes a problem

    from the scientific method perspective, where repeatability is normally seen as part of verification /

    validation. However, this is only a major limitation in the cases where we look for one particular solution,

    e.g., an optimum. In the intermediate phase involving clustering, we do not involve heuristics, but we

    have only considered the UPGMA clustering algorithm and the similarity coefficient of Gower, see

    Section 4.1. Nevertheless, the consequence is the same as for applying heuristics; we cannot claim

    optimality for our proposed solution(s).

    Technology Diffusion – Limitations

    We use Technology Diffusion to discount (i.e., reduce) the anticipated benefits of 

    implementing new technology / processes / etc., to reflect the true benefits during a transition period

    where an organization learns how to utilize the changes to their full potential. The Technology Diffusion

    operator (see Section 4.2.1) takes learning rate and initial skill levels as input. The learning rate is linked

    to the speed with which the organization is able to acquire skills, while the initial skill level refers to the

    amount of skills that are directly transferable to the new situation.

    The learning rate and initial skill levels are not trivial to estimate, and we have used heuristics

    in this research based on comparing the situation at hand with a known benchmark situation in terms of 

    the time it takes to reach a certain level of skills. Further, we have only considered Technology Diffusion

    as given in (Silverberg, Dosi et al. 1988; Silverberg 1991), and hence, there may be other and better ways

    of modeling learning for the purpose of evaluating the true benefits of changing to new processes. Thus,

    optimality cannot be claimed.

    c-DSP – Limitations

    We use the compromise DSP as a construct to organize our multi-objective model. Solving a

    multi-objective model implies finding some sort of compromise between the various objectives. To help

  • 8/18/2019 FPSO Chapter Viii

    26/36

    CHAPTER VIII

    Closure

     

    311

    find a compromise that comes closest to satisfying all the goals, we use scenarios that are based on non-

    formalized heuristics. Hence, a solution found by one designer may not be the same as the solution found

     by another, everything else being equal, i.e., the input to the method is person dependent, which affects

    repeatability.

    Further, we have not considered alternative ways to organize and solve our model. Hence, we

    cannot claim optimality in terms of the solving correctness nor the solving speed. The example problem

    and the case study represent discrete ‘optimization’ problems and are small enough to allow enumeration.

    Hence, we have not solved these problems by means of an optimization routine, which a bigger problem

    may require. In such a situation, we do not know how the compromise DSP would work, especially if we

    have to use DSIDES, the computer infrastructure in which the compromise DSP normally is solved.

    The HPPRM Framework– Limitations

    Being an integrator of the previously elaborated constructs, the major limitation to the HPPRM

    in its present state is repeatability, stemming from the use of non-formalized heuristics. However, we try

    to alleviate this lack of repeatability by giving detailed examples of our calculations in Appendix C.

    Further, the lack of comparison of alternative constructs implies that optimality cannot be claimed, which

    makes the HPPRM a non-rigorous framework for realizing Hierarchical Product Platforms. However, we

     believe that rather having a solution that is optimal for a short time, if ever, we hope our solutions are

    robust and applicable as the world is changing, and when no longer applicable, that they can by adapted to

    the new situations in an affordable way. To say it with Robin Cooper (the father of Activity Based

    Costing); we prefer the solution to be “approximately right rather than exactly wrong.”

    When it comes to comparison of alternatives, we believe there is a right time for everything.

    The scope in this research has been firstly to identify what kind of products we should be seeking to

    facilitate reuse in a marginal field context (i.e., product platforms and product families). Secondly, to

    identify how to realize the identified products (i.e., the evolutionary approach). Thirdly, to develop a

    framework to facilitate the right realization of the right products (i.e., the HPPRM). Hence, the issue of 

    comparing the initially chosen constructs to other alternatives, is viewed as part of the next phase – 

    continuous improvement by changing the process.

    8.3.2 The HPPRM – Its Robustness

    The HPPRM is based on some principles, namely, clustering of existing data, modeling of 

    alternative processes, and solving based on the compromise DSP construct. What if we for some reasons

  • 8/18/2019 FPSO Chapter Viii

    27/36

    CHAPTER VIII

    Closure

     

    312

    cannot use one or more of these principles; will the HPPRM still hold as a framework?  We answer this

    question by first looking at the clustering.

    Clustering is a means of identifying a potential for standardization. As the pool of items to

    compare increases, and as the characters for which they are to be compared increases, the need for 

    clustering also increases. However, if there are other ways of identifying / quantifying similarities in order 

    to identify a potential for standardization, these may be interchanged with clustering at any time.

    Evaluating alternative processes is part of increasing the ‘adaptiveness’ of an organization.

    This is a preferred but not a crucial task. What is crucial, however, is the ability to link proposed

    standardization schemes to cost and schedule impacts, so as to evaluate their goodness’. If this

    information is not readily available, or if this information cannot be estimated, the usefulness of this

    approach becomes very limited. The reason why is that the HPPRM rests on the very assumption that

    standardization (i.e., reduce the number of options) will improve cost and schedule, while compromising

    operational performance. Hence, we would like to find the right compromise, and without any means of 

    comparing benefits with drawbacks, there is little point in doing this exercise.

    The compromise DSP construct is used to organize the problem in order to prepare it for 

    solution, i.e., organization and preparation for solving is the main issue. Hence, any construct that can

    facilitate this can be used in the HPPRM framework with the right amount of adjustments. In this context,

    the solution we seek is a Hierarchical Product Platform, i.e., the variables we want to FIND are the level

    of standardization at each assembly level and the processes that gives he best combination of cost,

    schedule and operational performance. If we for some reason want another solution, this is easily adjusted

    in the FIND and he MINIMIZE part of the compromise DSP. Based on this we assert that the HPPRM

    framework is quite robust when it comes to accommodating changes and / or alternative constructs.

    8.3.3 The HPPRM – Limitations to the Validation Process

    In Section 8.2 we have validated the HPPRM framework by means of the validation square

     process. In this process, the assumption that has the weakest base is the Empirical Performance Validity

    (the usefulness). The reason why, is that it is based on our subjective estimates of benefits and operational

    scenarios. In order to alleviate this situation we have used the available information to establish upper and

    lower limits for what is possible, and have tried to operate conservatively between these limits to make

    sure we are not painting a too optimistic picture. In that respect, we see that standardization improves

    schedule and increase initial CAPEX, as would be expected. However, we see that downstream

    conversions are where the true benefits take place, as would be expected too. (This is confirmed by the

  • 8/18/2019 FPSO Chapter Viii

    28/36

    CHAPTER VIII

    Closure

     

    313

    Lockheed strategy for the C-130 Hercules). Hence, being the weakest link in the validation process, it is

    still strong enough to support the overall validity.

    8.3.4 The HPPRM – Limitations to Implementation

    The HPPRM is developed for one particular industry, namely, the offshore oil and gas industry,

    and we now estimate what it will take to implement this framework in this very industry. Firstly, the

    information to be clustered has to be available in some, way, shape or form. The level of readiness will

    vary, and this will impact the usefulness of clustering in terms of saving time and money to reduce the

    combinatorial problem. In this research, the information has been taken off of drawings and manually put

    into a database. We believe that as design increases the usage of centralized databases, it should not be too

    difficult to get hold of the raw data from which aggregated characteristics can be derived.

    If clustering is chosen to identify potentials for standardization, we recommend using a

    commercial software package. From such software one gets all kinds of information about the goodness of 

    the clustering itself, and normally, different similarity coefficients are available to choose from. Hence, as

    long as the data is available, we don’t see a big problem in getting it clustered.

    Evaluating the clusters require experienced people, that can identify which parts are not

    compatible for standardization, e.g., ship side and double bottom as discussed in Section 6.6.7. This

    assumption is valid in the case of Kvaerner, but may not be valid in other cases. However, analyzing a

     portfolio of existing designs indicates that there is. Over time, the expertise held by the experienced

     persons can be transformed into formalized heuristics and computerized, increasing the automation level.

    Regarding the latter, estimating learning rates and initial skill levels is a challenge, and will

    constitute a problem that has to be addressed. The important thing, however, is that we advocate to

    continuously look for better ways of executing a project. If alternatives are proposed, the effect of 

    implementing them should get attention. Furthermore, by demonstrating that the learning rate has an

    impact, we hope to emphasize that design is truly human-centered, and that continuous learning at the

    individual level may be the single most important thing that can improve an organization’s performance.

    In addition, by demonstrating that the initial skill levels also has a substantial impact, we hope to

    encourage small and frequent steps rather than few and large ones, i.e., we advocate evolution over 

    revolution.

    Regarding modeling and solving of problems in industry, this does not constitute a dramatic

    departure from existing practices. The type of problem may be different, and the input information may

    have to be generated, hence, consuming resources. Further, the solving scheme may not have an adequate

    infrastructure (discrete optimization problem), and the interpretation of the results may have to be learnt.

  • 8/18/2019 FPSO Chapter Viii

    29/36

    CHAPTER VIII

    Closure

     

    314

    What is important, however, is that if a company finds is worth while to pursue this strategy, they may do

    so without having to dramatically change their ways – it pretty much boils down to believing in the

     benefits.

    Finally, the HPPRM is not to be constantly applied, but at certain points in time where the

     product portfolio proliferates due to introduction of new family members and / or sub-systems addressing

    completely new applications, etc. In these instances, Path-Dependence is still applicable and the task 

     becomes to look at the new items and compare them to the standardized ones.

    Based on this, we are ready to recommend some avenues for future work that in our opinion

    will improve this framework.

  • 8/18/2019 FPSO Chapter Viii

    30/36

    CHAPTER VIII

    Closure

     

    315

    8.4 RECOMMENDATIONS –

    AVENUES FOR FUTURE WORK

    Having critically evaluated the research presented in this dissertation, our objective in this

    section is to identify some fruitful ways of expanding this work. Since this research is conducted in an

    academic setting based on an industrial problem, we see the road forking when it comes to future

    research. One avenue goes into industry dealing with implementation and further validation of our 

    research, and one goes into academia dealing with further refinement of domain independent issues risen

     by our research, see Figure 8-8. Hence, in Section 8.4.1 and 8.4.2 we examine the industrial and academic

    avenues respectively.

    8.4.1 Industrializing the HPPRM – the Industrial Avenue

    Further validation of the HPPRM framework is linked to a successful industrial

    implementation; first in one industry, then in other industries. However, we only make recommendations

    regarding how to implement the HPPRM in the offshore oil and gas industry. We still assume that the

    objective is to find the level of standardization and the corresponding process(es) that gives the best

    combination of cost, schedule and operational performance. In this context

    §  we recommend to revisit what to compare for similarities; there might be something more

     promising than plates, stiffeners, girders, etc. For example, maybe topside is more interesting

    from a standardization perspective than the hull.

    §  Having established what to compare, we recommend to look at how to make the required input

    readily available. This is directly affecting how to design and how to document the design.

    §  Having clustered the input, we recommend to look at meaningful interpretations of the

    clustered data from an industrial perspective. This should be linked to clustering research in

    academia, especially to the formalization of how to construct taxa, see Section 8.4.2.

    §  Regarding process partitioning, we recommend to identify which parts of the realization

     process(es) are more interesting from an improvement perspective. This should be linked to

    research regarding how to represent process partitioning in academia, see Section 8.4.2.

    §  Regarding the standardization itself, we recommend to formalize a set of heuristics in order to

    comply with ISO 9000 requirements regarding tractability of design decisions.

    §  Regarding modeling, we recommend evaluating the availability of required information. If not

    readily available we propose to change either the modeling or the performance variables. This

    should be linked to research regarding model representations in academia, see Section 8.4.2.

  • 8/18/2019 FPSO Chapter Viii

    31/36

    CHAPTER VIII

    Closure

     

    316

    §  Regarding solving, we recommend to research what scenarios represent our customers best;

    what objective function(s) addresses the customer requirements best; and what solutions

    (robust / optimized) gives the best overall economy. This should be linked to research in

    academia regarding solving algorithms, see Section 8.4.2.

    THE HPPRM FRAMEWORK --

    BASIS FOR SUGGESTING

    FUTURE RESEARCH

    Establish

    Scenarios

    Establish

    Scenarios

    THE ACADEMIC

    RESEARCH AVENUE

    (domain independent)

    THE INDUSTRIAL

    RESEARCH AVENUE

    (domain dependent)

    What to standardize

    Information availability

    What to partition

    Info availabilityMetrics for evaluation

    What to model

    Info availabilityMetrics for evaluation

    Mapping of customers

    - weighting schemes

    - operational scenarios

    Decision criteria

    - robust / optimized

    How to cluster 

    - similarity coefficients

    - clustering algorithms

    - cluster representation

    - construction of taxa

    How to partition

    - division criteria- metrics

    How to model:

    - cost/time/performance- learning

    How to solve- discrete problems

    - integer problems

    Research validation

    Metrics for evaluation

    What to solve for - objective functions

    Formalize heuristics

    Gather DataGather Data

    Standardize

    Designs

    Standardize

    Designs

    Decide on HPPDecide on HPP

    Solve for eachScenario

    Solve for each

    Scenario

    PartitionProcesses

    Partition

    Processes

    Model

    Relationships

    Model

    Relationships

    Cluster DataCluster Data

  • 8/18/2019 FPSO Chapter Viii

    32/36

    CHAPTER VIII

    Closure

     

    317

    Figure 8-8

    The Research Avenues Stemming from the HPPRM Framework 

  • 8/18/2019 FPSO Chapter Viii

    33/36

    CHAPTER VIII

    Closure

     

    318

    8.4.2 Improving the HPPRM Constructs – the Academic Avenue

    Having presented some recommendations regarding how to implement the HPPRM in the

    offshore oil and gas industry, we now present the corresponding domain independent research aspects. All

    of these aspects should be researched in parallel with research in industry in order harmonize the

    WHAT’s and the HOW’s regarding the HPPRM framework. Based on this we elaborate on some of the

    generalized concepts stemming from the HPPRM framework, see Figure 8-8 and Section 8.4.1.

    Clustering – How to Improve the Identification of Standardization Potentials

    Clustering is used to identify potentials for standardization by quantifying similarities. In our 

    research we have only considered the UPGMA clustering algorithm, the similarity coefficient of Gower 

    and the dendrogram representation, see Section 4.1. We recommend to evaluate whether other alternatives

    are better suited for the purpose of identifying potentials for standardization, i.e., to evaluate which

    alternatives are better for which situations.

    Having a clustering, the next thing is construction of taxa. In our research we did this by trial

    and error, applying our experience to evaluate when a good partitioning appeared. We recommend to look 

    into how to formalize a set of heuristics in order to computerize this process.

    Process Partitioning – How to Represent a Process

    Process partitioning is introduced to identify where the process can be improved in general,

    and where standardization creates new opportunities for process improvements in particular. In our 

    research we partition the process along a timeline, i.e., design, fabrication, installation, operation and

    maintenance, etc. From the perspective of improving the processes, there might be other partition schemes

    and we recommend looking into this. On the same token, metrics for evaluating how well a partitioning

    suits a given purpose also needs attention, and we recommend this as a future research area as well.

    Modeling – Revealing the Performance of the Proposed Standardization Schemes

    The purpose of modeling is to investigate the impact of standardization on cost, schedule and

    operational performance. In this research, we estimate deviations from a base-case for the various

    standardization schemes. However, this can only be done where detailed accounts for a base case exists,

    an assumption that rarely holds in academia. Hence, we recommend to look for other indicators that can

     be used to quantify how standardized systems compare to non-standardized systems in terms of cost, time

    and operational performance.

  • 8/18/2019 FPSO Chapter Viii

    34/36

    CHAPTER VIII

    Closure

     

    319

    In particular, we recommend to look at learning within an organization, not only from a

     process change perspective, but from a more general adapt-to-change perspective. In this respect, we

    recommend to look at ways of estimating learning rates and initial skill levels; what organizational

    characteristics are good indicators of learning rate, and what are good indicators of how much can be

    leveraged from current processes to new processes.

    Solving – Arriving at Acceptable Solutions

    The purpose of solving is to find the set of free variables that give the best performance for the

    system as described by an objective function. In this research, we have used the compromise DSP

    construct to facilitate solution, and we have solved by enumeration for both the example problem and the

    case study. However, we foresee situations where a search approach may be more appropriate, and we

    recommend further research regarding implementation of search algorithms into the HPPRM. In this

    respect, gradient based search does not seem too promising considering the nature of the discrete problem.

    Hence, we have indicated Genetic Algorithms or the Foraging ALP as given in (Lewis 1996) as

     promising starting points.

    Validation – Supporting the Decisions

    The purpose of validation is to build confidence that we make the right decisions, whether it is

    final decision on a standardization scheme or on implementing alternative processes. In this research, we

    have introduced the validation square, but the final step requires a ‘leap of faith’. We recommend to look 

    into how this leap can be made even shorter, i.e., how can we improve documentation of usefulness with

    respect to a purpose. We have suggested documentation by means of example problems and case studies,

    however, in academia example problems and / or case studies often become hypothetical and / or trivial.

    These potential avenues for further research are just a sample of the realm of topics waiting to

     be addressed. It is our hope that at least some of these topics are indeed taken on by researcher’s to come,

    so that this work may live on. Based on this we proceed to give a concluding remark in Section 8.5 to sum

    up the whole ‘shebang’.

  • 8/18/2019 FPSO Chapter Viii

    35/36

    CHAPTER VIII

    Closure

     

    320

    8.5 CONCLUDING REMARKS

    Before we conclude we look to Bob Mandel’s words:

    “Every Problem Was Once A Solut ion To A Previous Problem ” 

    - Bob Mandel

    We certainly hope that the HPPRM is not an end in itself, but a stepping stone for future

    research work in many fields of engineering – that we have created future problems for future researchers.

    We belive that the evolutionary approach shows prospects of providing an overall perspective on design

    that can help designers from different disiciplienes come together to ‘co-evolve’ products rather than sub-

    optimizing them.

    With this we conclude this dissertation to the words of Krishnamurti:

    “The wor ld wi l l change only w hen we have changed ourse lves and 

      – mo st signi f icant ly – abandon ed our resis tance to change” Krishnamurti

  • 8/18/2019 FPSO Chapter Viii

    36/36

    REFERENCES

    Bailey, R. (1997). The Design of Industrial Ecosystems. GWW School of Mechanical Engineering.

    Atlanta, Georgia Institute of Technology.

    Lewis, K. (1996). An Algorithm for Integrated Subsystem Embodiment and System Synthesis.

    Mechanical Engineering. Atlanta, Georgia Institute of Technology: 329.

    Mistree, F., O. F. Hughes and B. A. Bras (1993). The Compromise Decision Support Problem and the

    Linear Adaptive Programming Algorithm. Structural Optimization: Status and Promise, Washington D.C.

    Richardson, G. P. and A. L. Pugh (1988). Introduction to System Dynamics Modeling with DYNAMO.

    Cambridge, MA, the MIT Press.

    Silverberg, G. (1991). “Adoption and Diffusion of Technology as a Collective Evolutionary Process.”

    Technological Forecasting and Social Change 39: 67-80.

    Silverberg, G., G. Dosi and L. Orsenigo (1988). “Innovation, diversity, and diffusion: a self-organisation

    The Economic Journal 98: 1032-1354.

    Sneath, P. H. A. and R. R. Sokal (1973). Numerical Taxonomy. San Francisco, W.H. Freeman and

    Company.