The Dependence On Experimental Data

Published Date: 02 Nov 2017

Engineering research is moving away from being a craft towards being an engineering discipline, and as such requires good engineering practice just like any other branch of engineering. The larger and more complex a research project, the greater the need to impose discipline on the technical research activities.

Engineering is about design, so it is appropriate to follow the same procedure to design the products of engineering research in the same manner as in the design of any other engineered product. For the purposes of this procedure engineering research produces a set of products that serve particular purposes and perform particular functions. It includes the project management documentation that specifies how a research project can be completed on time and within budget.

This procedure identifies the research products and the processes needed to produce them. For engineered products beyond a certain complexity their quality can only be assured by producing them according to a well-defined engineering process that ensures quality products are developed and accepted in an orderly manner.

Engineering research is an innovative process and requires creative and flexible approaches rather than restrictive practices. Hence, this procedure follows a research strategy rather than enforcing a standardized methodology. The strategy is based on the production of clearly defined documentation, and is aimed at reducing the effort needed to get the documentation into a useable form.

In planning how to develop the research, it is the responsibility of the project management to ensure that a coherent system of methods and tools is chosen, integrated and supported. Differences in organization, applications and existing approaches, however, make it impractical to prescribe a single scheme that can be universally followed. Methods, tools, management practices or any other element of the total research environment cannot be chosen without considering each element in its relationship to the other parts of the environment.

The aim of this research procedure is to ensure that at each stage of the research process, the user, researcher and supervisor are certain of the correctness of the activities so far completed. The terms �user�, �researcher� and �supervisor� are used frequently throughout this procedure. They should be understood in a generalized sense in the context of their use, the user(s) using or requesting research from the researcher(s), to a quality ensured by the supervisor(s). The term �quality� is understood to be fitness for purpose at an economic cost.

This set of procedures have been produced in the interest of researchers with the intention of helping them to achieve high quality results. The document is a compendium of strategies, recommendations, checklists and advice. It can never be complete, and rigid adherence to its provisions cannot guarantee valid research.

The aim is that students become aware of:

� what engineering research is and why it is different from other research,

� why engineering research is guided by a research procedure, and

� the products of each phase of research.

The objective is that the student will:

� be aware of the quality control for each product of research,

� know the advantages of planning a research project, and

� understand the need to manage the progress of the research.

2.2 Research process

Before introducing the research procedure we need to find out what is involved in research. The activities involved and the products produced can then be identified and ordered to ensure that the research progresses in an efficient and satisfactory manner.

2.1.1 Research � brief summary

Research is essentially an experimental method for finding the behaviour of a real problem by modelling the real world behaviour with a finite number of variables. In this model the real problem is represented by a set of physical, mathematical or logical elements bounded by an idealized environment. Each element is assumed to be defined by its boundary, by its properties and by its assumed behaviour. The behaviour of the complete, idealized model is determined as the aggregate behaviour of its elements in relation to each other element.

Thus, any results obtained from the model are only as good as:

� the model of the problem,

� the assumptions embedded in the properties and behaviour of each element, and

� the representation of the assumed boundary conditions.

2.1.2 Planning the research

Before commencing any substantial research the researcher should prepare an outline plan relating the objectives of the research to the facilities, resources and time-scale available.

Aspects of the plan which may be relevant are related to the following key issues:

Aspects Resources Elapsed time Other requirements

Person days Cost

Familiarisation - with methods

- with facilities Consultation?

Literature survey?

Understanding problem

- results required

- level of detail Feasibility study required?

Problem formulation � data preparation

- data entry

- validation Modelling facilities?

Data acquisition � structure, properties

- boundary conditions

- interfaces Someone else�s work schedule?

Processing - model operation

- output presentation Model availability?

Interpretation - assessment

- real problem behaviour Analysis and presentation facilities?

The initial planning produces a Research Proposal and Terms of Reference.

2.1.3 Resourcing and authorisation

In general the researcher will be faced with one of two situations:

� an instruction to perform the research (presupposing supervisor�s awareness of probable cost/elapsed time), or

� a request for research, subject to cost or time-scale constraints.

In both cases the researcher should ensure that sufficient resources will be available at the right time to complete the research.

In the first case this may mean fitting in with other peoples work schedules, resource availability and facility access times, and conflict with other tasks for the researcher. The supervisor should always be made aware of implications before the research starts; i.e. perform a feasibility study.

In the second case the need for resource confirmation and authorization is obvious.

The feasibility study produces a Feasibility Report and usually an update of the Research Proposal and the Terms of Reference. When these have been accepted with approval to proceed, a research project should be established with a Project Manual.

2.1.4 Researcher and supervisor

Competent research can produce valuable results and increase the pool of knowledge. Incompetent research gives results which are, at best unreliable and, at worst, positively misleading. If safety depends on the results of the research, it should be performed, supervised or verified by an engineer of adequate experience and status.

There are no laid down standards for accrediting researchers and supervisors; the following guidelines should be followed:

� The researcher and/or supervisor should be educated to at least first degree standard in a relevant scientific discipline (engineering, physics, mathematics, computing, etc.).

� For degree purposes the supervisor should be educated to at least the level that the researcher will attain (e.g. for a Masters degree, the supervisor should have at least a Masters degree with a minimum of two years post-Masters experience).

� The researchers should be trained in problem modelling and use of research tools at their disposal, to the satisfaction of the supervisor.

� Training should either be formal, via reputable instruction courses such as this one, informal on-the-job via competent supervision or by reference to approved manuals and training aids.

� In the latter instance there should always be a fall back to an experienced and reputable expert for consultation.

� The supervisor should be capable of, and in all instances should conduct, an appraisal of research results based on practical appreciation and experience of related problems.

A researcher either has prior relevant experience, or has adequate professional training plus the ability, time and resources for familiarization with relevant facets of the research.

A consultant has extensive prior experience and high professional standing in understanding and applying research to relevant practical problems. It is important to establish the credentials of self-proclaimed experts, if possible by references.

A supervisor combines the normal qualifications for professional supervision with prior experience in the direction of similar research and/or expertise in the principles and practice of engineering research. The supervisor�s essential functions are to provide informed engineering guidance and critique of all aspects of a research specification and direction about the results interpretation.

2.2 Research specification

A clear statement of intent is helpful to the researcher and indispensable to any second party who will execute, use or contribute to the research. This is best embodied in a formal specification.

2.2.1 Problem description

It is always good practice to begin by showing drawings or sketches illustrating the problem to be solved. The researcher should then summarize the pertinent characteristics of the problem to be solved, using the following as a guide.

a The region of the problem to be researched.

� basic logic, topology, geometry or physical definition

� behaviour, material and construction

� special features

� adjoining components

� boundary conditions

� purpose of research

� sources of authentic data

b Nature of problem to be solved.

� linear or nonlinear

� static or dynamic

� fact or opinion

� general or specific

� animate or inanimate

c Nature of external effects � sources of authentic data

� real or artificial

� intrinsic or imposed

� fixed or free

� historical or to be collected

� environmental conditions

2.2.2 Model representation

Model representation deals with the researcher�s intentions for representing the research problem. Pertinent features, supported by brief justification, should be defined using the following as a guide.

a Diagrammatic

� boundary approximations

� subdivision and interfacing definition

� rules/algorithms for representing each region

� generation and graphical aids

b Tool selection

� tools used for particular regions of the model

� general or special representation

� modelling aids to be used

c Sub-scale features

The researcher will often be faced with detailed physical features whose scale is too small to be represented explicitly. In some cases these only have a very local effect but in others they significantly affect the behaviour in their neighbourhood. The method of allowance for such features (often very important in relation to the accuracy of results) should be clearly defined, using diagrams where appropriate.

d Symmetry and boundary conditions

Symmetry in the problem area can often significantly reduce the amount of effort needed if full advantage is taken by reflection across the boundary of symmetry. But be careful that proper boundary conditions are imposed.

e Data requirements

� fixed data as part of the model

� variable data for input to the model (modelling)

2.2.3 Results

The researcher must specify clearly what results are required, how they are to be selected and presented and what checks are required to validate them.

? Types of results

� physical

� numerical

� mathematical

� logical

� graphical

� textual

? Selection

� tabulation

� selection by cases

� selection by comparative criteria

� selection to suit codes or limits

� requirements from user

� interpolation and extrapolation

� selection aids to be used

? Presentation

� text display and tabulation

� direct point-by-point plots

� time history plots

� section plots

� contour plots

� interactive graphic displays

� bar or pie charts

� animated displays

� special presentation

� demonstration

� presentation aids to be used

? Diagnostics

� automatic checks

� special checks and criteria

� error/accuracy

� special representation and diagnostic aids to be used

2.2.4 Modelling methods

For many research problems there may be no suitable alternative methods available, for example survey questionaries, and hence no need for elaborate specification. However, in more complex cases, there may be choices for the researcher.

A theoretical or numerical model may depend for some of its data, or may be supported by physical tests. Any such conditions should be recorded at the outset.

? Dependence on experimental data

This situation should be handled with extreme care and in all but the simplest cases, should not be tackled without expert advice. Test data will always contain experimental errors which, in many cases, imply inconsistency with the assumptions embedded in the model. This is especially true if redundant data (e.g. more than one measurement of a single dimension) are determined independently by test. The following principles should be applied:

� Use test data, wherever possible, to determine �natural� physical properties, i.e. properties expressed in terms of the minimum number of fundamental physical variables (e.g. material stiffness expressed in terms of stress/strain or relative displacements).

� Use any redundant test data to smooth the natural properties by standard error-reduction methods.

� Expand �natural� data to conventional representation used in the model.

� In extreme situations, use simple devices such as taking the mean when that is physically demanded. Note that physical measurements are never as accurate as numerical results

? Theory confirmed by test

The researcher should try to ensure that the following match as nearly as possible, between theory and experiment

� physical geometry (especially tolerances), adjoining rig, etc.

� loading application

� supports, constraints and boundary conditions

� material properties and/or environmental conditions

� measurement/calculation location points.

2.3 Method validation

The basic theory and the model used for any research must be adequate to represent the problem, mathematically, numerically and physically robust in dealing with the many variations and features likely to be encountered.

Before commencing any significant research the researcher and supervisor should satisfy themselves that:

� The basic theory is sound and applicable to the particular problem.

� The algorithms and methods of application have satisfied certain fundamental tests for soundness, stability and convergence.

� Benchmark tests have been performed to demonstrate satisfactory performance.

� In special circumstances, specific tests have demonstrated performance in similar circumstances to the proposed research.

2.3.1 Basic theory

The researcher should ensure that the supporting documents and references provide adequate justification of the soundness of the basic theory. Since numerical approximations are often buried deeply within the theory, it is often difficult to establish integrity on a definitive, analytical basis.

2.3.2 Fundamental tests

Where complete analytical demonstration of integrity is infeasible, the modelling should be shown to satisfy a number of fundamental tests, specific to the nature and theory pertaining to the model.

First rate research will contain such justification within its own documentation, presented with clarity and conviction. If such documentation does not exist, the research should be treated with caution unless and until such tests are performed to the supervisor�s satisfaction.

2.3.3 Benchmark tests

Benchmark or acceptance tests serve many useful purposes, e.g. demonstration of performance, accuracy and reliability of basic theory and modelling facilities, indication of cost/time performance, familiarization with the model, its input/output and operational features.

Where major safety considerations arise or where unusual applications of a method are involved, it is advisable to seek specific confirmation of the method validity or accuracy.

This means relating the subject problem to a reference analysis which is either similar but simpler and backed by exact analytical or an established semi-analytical solution, or is similar in nature and complexity and backed by experimental evidence.

It is good practice to devise a specific and closely related experiment and perform a reference analysis for the express purpose of method confirmation.

It is also good practice, before embarking on any complex research to carry out a coarse analysis or grossly simplified analysis first, in order to:

� gain experience in the use of all facilities employed

� identify the critical regions and scale of idealization needed

� help in the selection of results.

2.4 Modelling and formulation

Selection and preparation of a model can have more bearing on the accuracy of results than choice of the basic methods and tools to be used.

The researcher must understand the tools being used and must be accurate and consistent in preparing data.

Automatic data checking facilities should be used as far as possible to verify that data are logically correct, physically meaningful and what the researcher intended.

2.4.1 Good modelling practice

� Always try to make the model represent all potentially effective material in the problem area.

� Traditional modelling practice was to simplify the problem by assuming primary solution paths and neglecting secondary effects. With powerful modelling tools this practice is usually unnecessary and always misleading. It should be reserved only for preliminary or supplementary analysis.

� Concentrate modelling detail in the regions of most concern.

� When two or more complex regions connect over a relatively small common boundary, structure the problem into sub-regions for economy in analysis and ease of definition.

� Subdivide complex problems into identifiable zones for modelling and checking purposes.

2.4.2 Geometry

All physical problems have geometric definition. The problem is normally represented by bounding lines and surfaces which represent:

� outer/inner boundaries

� centre lines/surfaces

� continuity/discontinuity

� connectivity

� symmetry

� constraints/releases

� boundary conditions

Nodal geometry is the most readily controlled of all data. Numerical accuracy of nodal geometry is important only up to a limited number of significant figures, usually compatible with the discrimination of the human eye.

The nodal geometry defines the physical shape and this may have a significant bearing on the accuracy of the modelling.

In general the best checks for geometry are graphical displays. Every significant geometric representation should use drawings, plots or computer screen displays for checking line and surface shape, continuity and connectivity.

2.4.3 Idealization

The researcher must carefully select the types of representation of the research problem best suited to the solution of the particular problem in hand and must specify behavioural properties which faithfully represent the real problem.

Only in the simplest cases is the choice of properties obvious and necessarily accurate, such as representing a thin surface by a solid or a thin shell. The following is a selection of typical situations confronting the researcher which affects idealization:

� Simplification of boundaries (complex curves replaced by straight lines, parabolas, conics, cubics) affects also the behaviour idealization.

� Particle size or perforations below analysis scale.

� Finite thickness of two dimensional area.

� Laminated construction represented homogeneously.

� Linear or nonlinear behaviour.

� Material properties, linear, nonlinear or time dependent.

� Idealized intersections and connectivity.

� Holes.

� Continuity.

2.4.4 External actions

External actions, forcing functions and environmental conditions provide the loading on the model whose behaviour is under consideration. These are often defined empirically, by code or by experiment outside the control of the researcher. It is important that data so supplied should be consistent with the physical geometry of the model.

Always record the source and standard of externally supplied data and demonstrate compatibility with the model. External actions are notoriously difficult to apply to the model such that they can be displayed graphically and applied in the correct format. It is necessary to take particular care in their preparation and this is one area where independent checking and careful engineering supervision are strongly recommended.

2.4.5 Boundary conditions

Boundary conditions, constraints and releases are probably the least understood and worst documented aspect of research models. They are such that they represent everything outside the model and hence should be established and formulated with extreme care. More often than not they are included as idealizations that over-simplify the real problem.

If the adjoining environment is passive (i.e. merely responds to the model behaviour) the boundary representation is usually fairly straightforward. Where there is interaction in both directions with the environment it is necessary to represent this interaction realistically at the boundary, or extend the model boundary to accommodate these effects.

For symmetrical models, it is often convenient to simplify the model by reflecting the behaviour across the symmetrical boundary. Likewise multiple symmetries, radial symmetry, cyclical repetition may be handled by using individual segments. It is good practice to perform trial transformations using a greatly simplified model with similar symmetry features to confirm that the correct representation is used.

Special care must be exercised in handling symmetrical models as there may be no reason for the external actions to show the same degree of symmetry as the geometry.

2.5 Solution methods or model processing

With any research problem there will often be a choice of methods of solution. If a decision has to be taken between alternative systems, the following may help selection:

� Physical models are usually unnecessary if a mathematical representation of the behaviour has already been established.

� Methods of solution with good error checking and diagnostics will provide more reliable results.

� Efficient solution methods are more economical if multiple runs are required.

� Accuracy is limited where ill-conditioning and singularities are present in the model.

The principal selection criteria will normally depend on economics (accuracy:cost) and the quality of the results when the limits of the solution method are approached.

2.5.1 Linear behaviour

Linear solution methods are normally stable and lead to few problems where the behaviour remains within the problem limitations. Care must be taken to ensure that the results are not physically unrealistic.

2.5.2 Nonlinear behaviour

Nonlinear behaviour is normally modelled in an incremental manner where the solution follows some prescribed nonlinear function. It is essential that the researcher clearly understands the natural problem and the assumed nonlinear behaviour, distinguishing between linear and nonlinear components. Complex physical instabilities may occur where the incremental steps are too large, and numerical instabilities with large processing times where the incremental steps are too small. However, in most cases straightforward and useful solutions can be obtained, by careful selection of the step sizes, modifying these at each step if necessary.

2.5.3 Discontinuous behaviour

Discontinuous behaviour leads to the most complex behaviour of all, insofar as solutions depend upon the history of the solution processing from a known datum state. The discontinuous function is normally a combination of linear and nonlinear relationships, with each discontinuity assumed to provide a new permanent state.

No known general purpose solution schemes are available for this regime or for identifying criteria for transition from stable to unstable behaviour. Small excursions into each new discontinuity are required to ascertain the stability and behaviour at each step, and the researcher is cautioned at attempting to model this type of behaviour.

2.5.4 Solution checking

Two main kinds of solution checks (apart from visual inspection for major formulation errors) are recommended as good practice:

� Checks for numerical solution accuracy and repeatability of results.

� Checks for compatibility of solutions with assumed nature of analysis and physical meaning.

2.6 Selection of results

Most modelling produces vast quantities of output that cannot be assimilated in total. This is particularly true when large numbers of external conditions are applied to the model, simultaneously or separately, and/or when the model is complex with a large number of boundary conditions. To aid assimilation, various kinds of selection should be specified, and these can be applied directly by the researcher.

The researcher should learn to be selective with output rather than specifying complete output as the latter are frequently dumped or filed without full appreciation of useful information. The needs of the research are paramount, but getting bogged down in a sea of output is disheartening.

The following are some of the selection options which may be available, or suitable for specification:

Selection criteria Typical result

Maximum parameter value for every point Highest one or more values with respect to many applied conditions.

Maximum or minimum derived characteristics. Characteristics derived from a function of several parameters.

Local parameter distribution in the neighbourhood of maxima. This gives the situation near a point for subsequent interpolation/extrapolation.

Local aberration in parameter values. Indication of ill-conditioning and singularities.

Maximum or minimum global parameter values. Minimum (or maximum negative) overall value of each parameter and maximum (or minimum negative) value of each parameter.

Direct output values above and below pre-specified limits. This locates critical values where they occur.

2.6.1 Presentation of results

The preferred methods of presenting results are:

? Graphical

� Distorted axis diagrams are extremely useful for understanding the model behaviour and also as diagnostic aids when checking for and locating errors.

� Animation may be valuable for visualizing transient results where the output is vast or for conveying the significance of the results; it adds greatly to clarity, if not information content, but may be expensive to produce in relation to benefits.

� Contour plots, particularly with coloured in-fill, are probably the most useful means of presenting variations in scalar quantities or absolute magnitudes of general quantities.

� Three dimensional plots should be able to be viewed from any angle, with provision for hidden line removal and sectional plots.

� �One picture is worth a thousand words� can be taken literally as a conservative estimate of the useful information content of a high resolution graphics display.

? Tabular

� Selected tabulations and numerical displays are useful for summarizing large amounts of output.

2.7 Interpretation of results

The researcher needs to interpret the results and present the findings in order to provide full value for the research. Interpretation should include:

� Results limitations

Inherent limitations, manual or technical report references, assumptions and modelling.

� Engineering assessment

Formal assessment of validity and justification of conclusions.

� Interpolation, extrapolation and smoothing

Continuous/discontinuous behaviour, curve fitting, interpolation/extrapolation points, checks and boundary compatibility.

� Real behaviour

Correction for assumed geometry, transformations based on assumptions, correlation with observation.

2.7.1 Results limitations

Modelling is essentially approximate and its results meaningful only within certain limits implicit in the basic assumptions. In general the analysis is numerically accurate as a solution of the formulated problem.

In a good model, the design will have made clear the limitations inherent in the results. Numerical results will normally follow smooth curves and surfaces capable of easy interpolation and limited extrapolation. They can be subject to substantial overall errors as a result of modelling limitations.

2.7.2 Engineering assessment

One of the commonest mistakes in applying any computing process is to assume that the output has the validity inherent in the processing accuracy of the software. This may lead, at best, to a complacency which may overlook errors or, at worst, to an inversion of logical reasoning to support an absurd result.

No significant results should be accepted without assessment to ensure their reasonableness. This assessment should, if possible, precede any complex processing of results.

Engineering assessment always involves the inspection of individual or combined results which can be presented in a familiar form for ease of comparison with prior experience, with global quantities or with trends in adjacent regions.

Specifically, a researcher may look at:

� Smoothness and physical form of result patterns (exaggerated scale plots are helpful).

� Significant resultants either globally, or at sections or subregions.

� Behaviour at boundaries and intersections.

� Contour plots.

� History plots.

� Trajectories of principal parameters.

It is not necessary for the researcher to have any specialist numerical or mathematical knowledge in order to make valid criticism of results; general experience of the physical problem should suffice.

Any query about the validity or meaningfulness of results must be answered by a credible physical explanation, using information from the modelling. No justification based solely on esoteric knowledge of the model should ever be accepted. However, when unexpected results are confirmed by sound supporting evidence, a good researcher will add this to the store of general engineering knowledge.

2.7.3 Interpolation, extrapolation and smoothing

Modelling, by definition, usually involves physical or numerical approximation to the real world. Continuity is often physically specified only over certain regions; it is important to identify continuous regions and to specify the type of discontinuity expected at boundaries.

Thus, before attempting any type of curve or surface fitting, divide the model, and the appropriate output data into continuous zones representing the physical reality.

Surface/curve fitting involves interpolation and may require regular spacing of data points. This can be quite inconvenient unless the model definition nodes were originally chosen with curve fitting in view. If the nodes are chosen for accuracy of definition which does not lend itself to curve fitting, significant errors may result.

There is no guarantee that interpolated/extrapolated results, anywhere other than at a node definition point, will be other than an approximation. Extrapolation is only really meaningful at points within a short distance of the definition points, and accuracy can fall away where there is large curvature.

2.7.4 Real behaviour

The process of idealization and solution selection replaces the real behaviour by an idealized model for research purposes.

Idealization involves a number of formal or implicit operations to determine physical equivalence which may be inverted or reversed when interpreting modelling results. Where there is a one-to-one equivalence between idealized and real components (e.g. when making small corrections for curvature) a reverse transformation is readily identified and will provide a simple result correction.

Very often this is not the case, e.g. where discrete components have been lumped together. In this situation there is no unique inverse and it is necessary to devise a transformation which is at least fully consistent with that used in idealization.

2.8 Reporting the findings

Every significant piece of research should be well documented making as much use as possible of computer generated documentation of actual data submitted and derived. Ideally the majority of formal documentation should be self-generated as a normal by-product of data entry and results presentation.

Full documentation should include:

? Problem specification.

? Supporting diagrams illustrating

� the real physical geometry

� the idealized geometry

� interfaces with neighbouring regions and the environment

� required results presentation

? Key input data including

� basic geometry data or data sources

� plots showing locations

� detailed definition of boundary conditions and constraints

� definition of external actions

� key drawing references including standards and codes

� idealization data, procedures and key parameters

� materials and environmental data

? Selected output data as defined in the specification and normally including

� processed geometry plots

� selected tabulations

� contour plots

� section plots

� history of selected results

� other results

The research is not complete until an interpretative report is produced. This is a step often neglected by researchers who tend to consider their research complete when results are published.

The researcher is usually the best person to interpret the results for others as after the results are produced the researcher is the current world expert on that particular problem. It is good practice to conclude any exploratory research with a brief report or paper containing the following:

� Purpose of the investigation.

� Outline of the model and problem presentation.

� Summary of principal results.

� Discussion of results especially physical significance and accuracy limits.

� Conclusions and recommendations.

2.9 Research strategy

Research has traditionally been conducted according to a well-established standardized methodology which tends to be inhibiting, inflexible and self-serving. Having determined what is involved in research we are now in a position to classify the various research activities and identify their research products. This will enable us to develop a research strategy aimed at reducing the effort needed to get quality products on time and efficiently utilizing the available resources.

For the purposes of this procedure engineering research produces a set of products that serve particular purposes and perform particular functions. It includes the project management documentation that specifies how a research project can be completed on time and within budget.

Research activity Research product

Getting started Title

Research proposal

Terms of reference

Feasibility study Feasibility report

Abstract sheet

Table of Contents

Project management Project manual

Research project

Literature survey Literature review

Preliminary model design Draft model design report

Draft theory manual

Detailed model design Model design report

Theory report

Model implementation Model documentation

Model operations manual

Model test problems

The model

Model testing Tested model

Assurance report

Modelling Input data

Output results

Modelling guide

Results processing Tables, graphs, charts, etc.

Research reporting Research report

Research publishing Research paper

Research presentation Oral report

Tasks: Hypothesis: �The moon is made of green cheese.�

1 How would you go about establishing that this statement is a fact?

2 How would you organize your research?

3 What steps would you take?

4 What resources would you need?

5 How long would it take?

6 How much would it cost?

Write a brief (less than two pages) proposal for you to undertake this research.