ACKNOWLEDGEMENTS
The author would like to express sincere thanks to Dr. Derek Carruthers at The University of Western Australia for his excellent supervision and enthusiastic assistance throughout the rather protracted development of this project. His respect and trust in this work over the past six years is probably the sole reason for its ultimate completion.

1 Abstract >>

---

2 Design Tools >>

• 2.1 Definition >>
• 2.2 Computer Modelling and Efficient Design >>
  • 2.2.1 Increasing Design Confidence >>
  • 2.2.2 The Validity of Design Tools >>
  • 2.2.3 Making a Difference >>
• 2.3 Building Performance >>
  • 2.3.1 Lighting Analysis and Simulation >>
  • 2.3.2 Shadow Analysis >>
  • 2.3.3 Thermal Performance Analysis >>
  • 2.3.4 Computational Fluid Dynamics >>
• 2.4 Towards Integration >>

---

3 A Conceptual Design Tool >>

• 3.1 Conceptual Design >>
• 3.2 Next Generation Design Tools >>
• 3.3 A New Approach >>
• 3.4 An Intelligent Interface >>
• 3.5 Interactive Modelling and Editability >>
  • 3.5.1 3D Cursor System >>
  • 3.5.2 Relationship Mapping >>
  • 3.5.3 Nodal Manipulation >>
• 3.6 Graphical Display of Inputs and Outputs. >>
  • 3.6.1 Establishing a 3D Viewpoint >>
  • 3.6.2 Working in 3D >>
  • 3.6.3 Displaying Calculations in 3D. >>
• 3.7 Multi-level Inputs >>
  • 3.7.1 Progressive Data Input >>
  • 3.7.2 Process Modelling >>
• 3.8 Use of Components and Library Data >>
  • 3.8.1 Material Libraries >>
• 3.9 Interoperability with Other Tools. >>

---

4 The Implementation >>

• 4.1 Automatic data gathering. >>
  • 4.1.1 Surface Areas >>
  • 4.1.2 Volumes >>
  • 4.1.3 Intersection Points >>
  • 4.1.4 Adjacency >>
• 4.2 Shadow Analysis >>
  • 4.2.1 Sunrise and Sunset >>
  • 4.2.2 Horizontal and Vertical Shadow Angles (HSA & VSA) >>
  • 4.2.3 Viewing From Sun >>
  • 4.2.4 Overshadowing >>
  • 4.2.5 Isolating Individual Shadows >>
  • 4.2.6 Shadow Profiles >>
• 4.3 Stereographic Analysis >>
• 4.4 Solar Exposure >>
  • 4.4.1 Global, Beam and Diffuse Irradiance >>
  • 4.4.2 Shading Factor >>
  • 4.4.3 Solar Exposure Graphs >>
  • 4.4.4 Angular-Dependant Reflectivity >>
• 4.5 Shading Device Design >>
  • 4.5.1 Solar Profiles >>
  • 4.5.2 Solar Extrusions >>
• 4.6 Daylight Factor >>
  • 4.6.1 Geometric Method >>
  • 4.6.2 Internally Reflected Component >>
  • 4.6.3 Translation Into Illumination >>
• 4.7 Artificial Lighting >>
• 4.8 RADIANCE Output >>
• 4.9 Thermal Performance >>
  • 4.9.1 Internal Temperatures >>
  • 4.9.2 Heating and Cooling Loads >>
  • 4.9.3 Passive Design Analysis >>
  • 4.9.4 Isolated Component Analysis >>
• 4.10 Statistical Acoustics >>
• 4.11 Acoustic Raytracing >>
  • 4.11.1 Sprayed Rays >>
  • 4.11.2 The Image Method >>
  • 4.11.3 The Hybrid Image Method >>
  • 4.11.4 Relating Room Geometry to Room Response >>
  • 4.11.5 Derived Information >>

---

5 Conclusion >>

---

Bibliography >>

Abstract

"Environmental design replaces structure as the principal problem of architectural science" [Cowan, 1966]. In response, more than 20 years later, Manning writes: "Despite enormous amounts of research that has been undertaken into many aspects of building environment, and the store of knowledge that has accumulated, design of the environment too often appears to be a matter of chance. Users of today’s new buildings are just as liable as were users of earlier buildings to be uncomfortable." [Manning, 1987].

A significant amount of the research referred to by Manning has been directed into the development of computer software for building simulation and performance analysis. A wide range of computational tools are now available and see relatively widespread use in both research and commercial applications.

The focus of development in this area has long been on the accurate simulation of fundamental physical processes, such as the mechanisms of heat flow though materials, turbulent air movement and the inter-reflection of light. The adequate description of boundary conditions for such calculations usually requires a very detailed mathematical model. This has tended to produce tools with a very engineering-oriented and solution-based approach.

Whilst becoming increasingly popular amongst building services engineers, there has been a relatively slow response to this technology amongst architects. There are some areas of the world, particularly the UK and Germany, where the use of such tools on larger projects is routine. However, this is almost exclusively during the latter stages of a project and usually for purposes of plant sizing or final design validation. The original conceptual work, building form and the selection of materials being the result of an aesthetic and intuitive process, sometimes based solely on precedent.

There is no argument that an experienced designer is capable of producing an excellent design in this way. However, not all building designers are experienced, and even fewer have a complete understanding of the fundamental physical processes involved in building performance. These processes can be complex and often highly inter-related, often even counter-intuitive.

It is the central argument of this thesis that the needs of the building designer are quite different from the needs of the building services engineer, and that existing building design and performance analysis tools poorly serve these needs.

It will be argued that the extensive quantitative input requirement in such tools acts to produce a psychological separation between the act of design and the act of analysis. At the conceptual stage, building geometry is fluid and subject to constant change, with solid quantitative information relatively scarce. Having to measure off surface areas or search out the emissivity of a particular material forces the designer to think mathematically at a time when they are thinking intuitively.

It is, however, at this intuitive stage that the greatest potential exists for performance efficiencies and environmental economies. The right orientation and fenestration choice can halve the air-conditioning requirement. Incorporating passive solar elements and natural ventilation pathways can eliminate it altogether. The building form can even be designed to provide shading using its own fabric, without any need for additional structure or applied shading. It is significantly more difficult and costly to retrofit these features at a later stage in a project’s development.

If the role of the design tool is to serve the design process, then a new approach is required to accommodate the conceptual phase. This thesis presents a number of ideas on what that approach may be, accompanied by some example software that demonstrates their implementation.

Design Tools

Definition

The term design tool is generally applied to a wide range of techniques, from the use of tabulated data sheets and manual calculation methods, through to sophisticated computer analysis software. In the context of this work, the term is used to describe computer software developed to replace laborious manual calculations used to inform the design decision-making process. Using a computer to perform the mathematical component makes it possible to study effects not previously considered in many building designs.

Computer Modelling and Efficient Design

Performance modelling and simulation are becoming increasingly important in building design, particularly during the building services phase. The forthcoming CIBSE Application Manual on energy and environmental modelling is set to endorse computer modelling as part of the engineering design process [Ruyssevelt 1997].

Whether driven by social conscience or purely financial concerns, there is a concurrent move towards better performing and more efficient buildings. Many designers see the increased use of modelling and simulation tools as a way of achieving this.

Increasing Design Confidence

Building design is full of uncertainty. Whether it is the architect uncertain of how much light a window might let in, or the mechanical engineer uncertain of exactly how much solar gain to allow for, this lack of surety results in a significant amount of over-design.

It is standard practice throughout the building industry to include safety factors and margins for error within design calculations. Studies have shown that these are generally between 5 and 10% [Parsloe 1995, Brittain 1996]. However, research by the Building Services Research and Information Association (BSRIA) suggests that such small margins, when applied many times at different points within a design, can result in a cumulative design margin of up to 60% [Race 1997].

There are many valid reasons for the use of safety factors, from uncertainties over the validity of a calculation method to the dubious accuracy of a manufacturer’s material specification. In some cases the information required in a design may be difficult to properly quantify, such as the variability of a material’s thermal resistance with moisture content. The spare capacity that results from over-design may even be desirable, allowing for future expansion or even deterioration in a system’s performance over time.

Whilst the use of design tools may not automatically eradicate high safety margins and over-design, experience shows that they can significantly increase a designers confidence in the solution [Palmer 1997]. Most design margins result from the perceived application of a generalised solution to a very specific problem. As the use of most advanced simulation tools begins with the entry of a very detailed model, the results produced at least appear to be more specific to the actual problem they are being applied to.

The Validity of Design Tools

The underlying assumption here is that results produced by a design tool are more accurate than manual calculations or rules of thumb. Some doubt does exist over the true validity of computer simulation methods when compared directly with real buildings [Bunn 1995], however, there is relatively little doubt about their improved accuracy over other calculation methods.

This has been shown in any number of projects where sophisticated thermal analysis and computational fluid dynamics (CFD) applications have been used. The increased number of design parameters that can be considered and the level of detail to which performance can be simulated is far beyond the scope of any manual method.

There is always potential for the inappropriate application of a such a design tool, or even the input of erroneous data. However, this is true of any method.

Making a Difference

Since the early 1960’s, the use of computer modelling and simulation tools within the building industry has steadily increased. These tools have progressed from simple single-task applications with limited input and output requirements [Howard 1960, Belchambers, et al 1961], to quite sophisticated modelling systems that can simultaneously analyse a range of performance parameters.

However, after nearly 30 years of development, there is still some scepticism as to the necessity and applicability of such tools to the design process. The basic question still remains: "Does the use of simulation and validation tools actually produce better buildings?".

The main focus in the development of design tools has been the accurate simulation of natural processes, heat flow through materials, the turbulent movement of air and the inter-reflection of light. The result is a range of software tools well suited to the task of detailed design validation. However, the user interfaces and specialist skills required to properly drive them mean that they are still very much part the engineering domain, used only in the final stages of a project.

The real challenge is to make these tools applicable to the earlier stages of design, where a more informed analysis of possible alternatives can yield the most benefit and the greatest cost savings, both economic and environmental. Specialist skills may still be required, but just as important is the ability to translate the loosest architectural sketch into a valid input model and translate the result into fundamentally solid design feedback.

This is where computer modelling and simulation can really lead to better and more efficient buildings, both in terms of their internal environment and the environmental impact on their surroundings.

Building Performance

The invention of the Carrier air-conditioning system is seen by many as having liberated the architect from the constraints of climate. The capability of mechanical services to produce a controllable and comfortable internal environment within any building is almost unquestioned in modern architecture.

This capability is well understood by architects and, together with artificial lighting technology, underpins the majority of building design. However, there is still a significant demand for non-mechanically serviced and naturally lit buildings as the social and economic benefits of passive environmental controls gain recognition amongst both architects and clients.

Unfortunately, there is concern as to the finite capacity of natural systems, compared to the perceived infinite capacity of mechanical systems. This make designers wary of passive design solutions unless they can be rigorously validated and clearly shown to work in each specific case. This is where modelling plays its part, making it more viable to use windows and vents instead of electric lights and air-conditioning by providing some assurance that the same levels of occupant comfort can be achieved.

The rest of this chapter is a brief analysis of exactly what performance characteristics can be modelled using existing design tools and how those results can be used to aid building design.

Lighting Analysis and Simulation

The aim of any form of lighting simulation is to be able to predict lighting levels at points within a space or over an entire surface. The process of accurately calculating these levels can become quite involved given the complex nature of surface inter-reflection. As a result, there exist a range of methods of varying sophistication for lighting analysis and simulation.

The simplest means of calculating light levels is to simply sum the contribution of each visible light source. This is known as the point-by-point method and takes account of the distribution pattern of each source. Indirect light and diffuse reflections off surrounding surfaces are not considered. This is a useful method for examining the distribution of light from multiple sources. The resulting levels significantly underestimate actual levels, however they offer important feedback at the design stage of a lighting system.

Calculating natural light levels is more difficult as diffuse light from the sky dome itself is often the main source, not just the direct sun. As cloud cover is a significant variable, a model of the distribution of light over the sky has to be used. The Commission Internationale de l’Eclairage (CIE) define a number of such models, for overcast, uniform and sunny conditions. These models were developed for use with the Daylight Factor method of daylight estimation, which is defined by CIE as follows:

‘The Daylight Factor is the ratio of the daylight illumination at a point on a given plane due to light received directly or indirectly from a sky of assumed or known luminance distribution, to the illumination on a horizontal plane due to an unobstructed hemisphere of this sky. Direct sunlight is excluded for both values of illumination.’ [Longmore 1968]

Daylight factors can be calculated manually using the British Research Station (BRS) Daylight Factor Protractors or lighting applications such as Adeline from the International Energy Agency (IEA).

More sophisticated methods of lighting analysis make use of full rendering techniques. The aim when using more advanced lighting simulation tools is to produce realistic images of the spaces within a geometric building model that correspond as closely as possible to what would actually be found if that space were real. There are many applications that produce rendered images from three dimensional models. However, there is a major difference between photo-realistic rendering and actual lighting simulation.

The following extract explains this difference quite succinctly. This is taken from notes written by Greg Ward to accompany the RADIANCE lighting simulation tool developed at Lawrence Berkeley Laboratories.

"Photo-realistic rendering places emphasis on the appearance of its output rather than the techniques used to derive it. Anything goes, basically, as long as the final image looks nice. There is no attempt to use physically realistic values for the light sources or the surface reflectances. In fact, the light sources themselves often have physically impossible characteristics like 1/r falloff (as opposed to 1/r²) or there is a lot of ambient lighting that comes from nowhere but somehow manages to illuminate the room. (You are probably saying "Hey! Doesn’t RADIANCE use an ambient term?" The answer is yes, but only as a final approximation of the intereflected component. The renderers I’m talking about use the ambient level as the main source of illumination!) Also surfaces typically have colour but there is no reflectance given, so all the surfaces appear to have the same brightness."

"Physically-based rendering, on the other hand, follows the physical behavior of light as closely as possible in an effort to *predict* what the final appearance of a design might be. This is not an artist’s conception any more, it is a numerical simulation. The light sources start in the calculation by emitting with a specific distribution, and the simulation computes the reflections between surfaces until the solution converges." [Ward 1994]

Light levels shown in images generated from detailed simulations of the physical processes of light are actual levels that can be read directly. This means that they have direct relevance to a particular design, indicating if levels are too low or if sources of glare discomfort are present. Post-processing of such image can be used to overlay contour lines or even to derive more complex information such as illuminance levels, daylight factors, glare indexes and even visual comfort.

The most sophisticated of lighting simulation techniques involves the analysis of flux transfer between surfaces. Known as "radiosity", this technique requires that all surfaces be divided into small patches that exchange light energy within a closed system. The computation is iterative, progressively refining the energy exchange until the variance between iterations is insignificant. Ashdown [1994] gives a comprehensive account of this technique and its practical application.

Radiosity is very computationally intensive. For the most part, it is limited to relatively simple scenes with mostly diffuse surfaces. Highly specular surfaces or very complex geometry can lead to excessive convergence times.

Hybrid techniques which combine radiosity with ray-tracing are used in some simulation tools. These involve tracing light rays in the reverse direction, from a point on the surface of an object back to its sources of direct and indirect light. RADIANCE is an example of a tool that uses a hybrid technique. It uses a geometric rays to determine the portion of other surfaces visible from sample points. Ambient light levels are then iteratively calculated for all points by interpolating between sample points.

A less accurate method is that of geometric ray-tracing. This technique is adequate for representing the effects of direct lighting and low order reflection, but does not represent high order intereflection or diffuse light well at all. This tends to be the technique used by most photo-realistic rendering applications used for artistic visualisation.

Shadow Analysis

An adjunct to lighting simulation is the analysis of sun penetration and solar shading. Whilst there are a number of dedicated shadow and reflection analysis applications available, such as SR [Owen and Roy, 1989] and SunCast [IES 1997], shadow analysis is mainly performed using CAD-based rendering and visualisation tools. Applications such as ArchiCAD and Microstation are examples of CAD packages with in-built solar position routines.

Many thermal analysis tools include functions for calculating and displaying shadows, such as FACET [IES 1997] and TAS [EDSL 1995].

Thermal Performance Analysis

The aim of thermal analysis is twofold:

• To accurately predict the internal temperatures of spaces within a model,
• To calculate heating and cooling loads.

Once again, there are a wide range of simulation techniques with varying levels of sophistication and accuracy.

Computational Fluid Dynamics

Computational Fluid Dynamics (CFD) is a technique used for the prediction of airflow patterns, temperature distribution and contaminant movement within buildings and other enclosed spaces. This is a very computationally intensive technique, with some models sometimes requiring several days to converge on a solution.

CFD is well suited to the assessment of air flow in complex spaces such as atria, malls and high heat load areas as well as modelling smoke propagation. Its main benefit in design is its ability to model air movement and temperature distribution in fine detail, allowing complex natural ventilation and displacement ventilation designs to be simulated and validated. This can lead to significant savings in plant and equipment, as well as greater levels of occupant comfort.

Towards Integration

There have been a number of attempts at some level of integration within these diverse areas of building performance. An excellent example of this is the Archipak software suite for the thermal design of buildings [Szokolay, 1987]. This integrated climatic analysis with solar geometry and a simple thermal analysis engine based on the admittance method.

More recently, ongoing development of the FACET suite includes acoustic analysis, electrical and mechanical system design, thermal simulation, psychrometrics and shadow analysis. Whilst this consists of a number of individual applications, consistent data models and file formats allows building information to be shared amongst them.

The 4D <Virtual Environmental> System being developed by Integrated Environmental Systems (IES) is another example of an attempt at high level integration. It includes a range of dynamic performance analysis tools linked to the one building model, the 4D-IDM file. These consist of visualisation and animation tools, terrain modelling and landform generation, CFD, fire and smoke modelling and egress analysis.

A number of organisations and associations have been formed to promote and facilitate this integration. The International Building Performance Simulators Association (IBPSA), as defined in their newsletter ibpsaNEWS, was established to "advance and promote the science of building performance simulation in order to improve the design, construction, operation and maintenance of new and existing buildings worldwide".

The International Alliance for interoperability (IAI) is another group formed by a number of construction companies and CAD software vendors with the aim of improving the inter-communication of building and project information. Founded initially in North America, there is now also a UK chapter. The main focus of their work so far appears to have been the establishment of Industry Foundation Classes (IFC), a set of standard object definitions and data structures for describing building elements.

The gradual integration and standardisation of design tools is a logical progression if performance analysis and simulation is to gain wider penetration and acceptance within the building industry.

The one application that appears to be taking a slightly different approach is the Building Design Advisor (BDA). This is being undertaken as part of the Building Technologies Program, Environmental Energy Technologies Division at Lawrence Berkeley National Laboratories.

This software integrates a number of simulation tools and databases, however, it attempts to accommodate the initial schematic phase of building design as well as the detailed specification of final components and systems [Papamichael 1997]. Whilst not likely to be available until late 1998, it represents a substantially new approach. An approach that recognises the designers role at the earliest stages of design.

This is the area least well served by existing design tools yet, it is argued here, with the greatest potential for their use. It is hoped that BDA and the software described in this work will go some way to filling this void.

A Conceptual Design Tool

Conceptual Design

The conceptual stage of design occurs very early in the design process. This is the time when a vast array of competing requirements are shaping the initial building form, when geometry, materials and orientation are still being formulated. As these are arguably the three most important determinants of building performance, this is the most crucial stage of a project.

Conceptual design is an iterative process of generating ideas that then need to be evaluated and tested, for rejection or further refinement. Traditional methods of testing an idea involve quick perspective sketches, simple geometric analysis on a drawing board, or even small hand-calculations. The main criteria for these tests is speed. Being able to quickly reject impractical ideas can save significant amounts of time. Each newly rejected idea providing one more clue to a more acceptable one.

A major part of this testing process is play - simply playing around or experimenting with an idea until it is shown to work or not. The purpose of this is to gain some understanding, both spatially and operationally, of the full requirements of the final form [Akin 1978]. Using traditional techniques, the range of testing is quite limited.

In order to make environmental performance a practical consideration at this early stage, thereby informing the decision-making process as much as any other consideration, real and useful feedback has to be produced from what is often ill-defined and abstract information. The precise and detailed input requirements of most existing design tools preclude this. To use them, the designer must first enter the small amount of hard data they do have, and then arbitrarily quantify whatever else is needed before a result can be produced.

Overcoming this requires a completely different approach from the concise, solution-based nature of existing analysis tools.

Next Generation Design Tools

In preparation for the development of a new energy simulation tool, based on an amalgamation of two existing programs BLAST and DOE-2, the US Department of Defence and Department of Energy co-sponsored two workshops on next-generation building energy simulation tools [Crawley, et al 1996].

The first workshop involved only developers and expert users of such applications whilst the second was open to general users. Participants were led through a series of creative brainstorming exercises designed to highlight areas requiring further development. The ideas produced were then collated and prioritised by vote to represent the group view.

In summarising the results of both workshops, the authors state: "Surprisingly, not many new or unusual ides were brought up - even with a group of building energy simulation developers and users. The hundreds of ideas generated during both workshops showed instead that the field of building energy simulation still has many fundamental problems that need to be addressed. The developers will not stretch the boundaries and capabilities of simulation until more basic simulation issues are resolved." [Crawley, et al 1997].

This is perhaps a pessimistic view that needs some qualification. Firstly, the context of both workshops was centred heavily on energy simulation and thermal analysis tools, and secondly, the participants of both workshops had significant exposure to the use of such tools. How much this affects their expectations of future tools is probably demonstrated by the workshop results.

Despite this, a number of very useful ideas were documented. In the following sections, a number of these ideas will be analysed and interpreted to form the substructure of a new conceptual design tool.

A New Approach

In order to be used at the earliest stages in the design of a building, any next-generation design tool must overcome the psychological separation between design and analysis that existing tools have created. As discussed previously, the primary cause of this is the detailed nature and amount of input required to describe a building model. Having to enter this data very early in the design acts to interrupt the process of iterative decision-making and forces the designer to prematurely make a series of arbitrary decisions just to produce a model acceptable to the tool.

A conceptual design tool must make the process of entering this data part of the design process itself. This is only possible if there are enough tangible benefits associated with having a model in such a format. The key to this is feedback, producing real and useful design feedback at every stage of the modelling process from data entry right through to final analysis. This places the focus firmly on the interface, the means by which the user describes and interacts with the model.

The aim is to have the creation of a simulation model replace, to some extent, the act of sketching. The primary role of the sketch is to assist the designer formalise ideas and test them against a range of design constraints [Akin 1978]. Therefore, to function similarly, the simulation model must represent building form and assist in the visualisation of the design. This advocates a very geometric approach to representation.

The challenge in this approach is to produce an interface within which geometric modelling is as simple and ‘loose’ as a sketch, yet can be used for detailed analysis at any stage in its construction. In addition, just as sketches develop and become more sophisticated, so too must the geometric model and its analysis. Further, a sketch can focus in on a very small area of a model or only one aspect of its nature. This must be true of the simulation tool as well.

An Intelligent Interface

Accommodating these requirements requires a very flexible interface, capable of satisfying the following requirements:

• Reducing the perceived input requirements to define a model,
• Maximum utilisation of whatever information is input,
• Allowance for constant development and refinement of the model.

• Interactive modelling and editability
• Full graphical display of inputs and outputs.
• Multi-level inputs for all calculations
• Simultaneous performance analysis
• Use of components and library data
• Interoperability with other tools

Interactive Modelling and Editability

The heart of any conceptual design tool must be its geometric modelling interface. As discussed previously, existing CAD interfaces are quite inappropriate to the inexact and interactive requirements of preliminary idea testing. At this stage, creating a model should be just about as simple as sketching it.

Sketches are usually quite simple and quick to produce. Several may be needed at different angles to test a particular idea. As such, they are disposable. The use of a computer sketching tool is only productive if:

• It is as quick and simple to create as traditional techniques,
• it can be used in place of a number of different sketches or
• it is editable and malleable enough to evolve along with an idea.

3D Cursor System

Very few designers sketch exclusively in two dimensions. Perspective views are an important tool for visualising spatial relationships between objects. As such, the ability to create objects within perspective views is considered very important. This has a number of ramifications for the type of cursor system to be used.

In applications such as AutoCAD and Microstation, the cursor is essentially two-dimensional. The orientation and offset of the working plane can be changed, but movement is restricted to this plane. Therefore, to create an object, the right cursor plane must be established first. Such a system does not allow for changes in this plane mid-creation or whilst an object is being moved.

There are a number of ways of implementing a fully 3D cursor. The method chosen here is to use keyboard modifiers to interactively alter the axis in which the cursor moves. This allows changes in height or direction at any time, even whilst dragging an object or node.

A 3D cursor position is determined from the 2D pointer by generating an imaginary ray travelling from the eye point through the picture plane at the pointer position and then through the cursor plane. Where this ray intersects the cursor plane is the 3D cursor position. The Shift and Control keys are used here to alter the cursor plane from the X-Y plane to the X-Z or Y-Z planes, located at the current cursor position.

Figure 3.1 - Interactively changing cursor plane.

To move a point up in Z whilst dragging using this technique, the user simply holds down the Shift or Control key, drags the 2D pointer up until the right height is reached, releases the modifier key and continues to drag the point into position. Experience has shown this to be an extremely intuitive method that most users pick up relatively quickly.

Object Snapping

In order to accurately relate objects spatially, the concept of snap points is widely used in many CAD packages. This allows a defined point on one object to be exactly aligned with a defined point on another. Object snapping provides a high degree of numerical accuracy even for objects created within a sketch-like environment.

The cursor snap mode can be set to any of the following mode at any time using either the mouse or keyboard:

• Off,
• Grid only,
• Nearest model node,
• Orthographic movement in one axis only
• Axial alignment with existing model nodes
• Intersection points between lines
• Mid-points of lines
• Object centre points

Restricted Movement

When working with objects that are geometrically related to others, there are times when their movement may be restricted. An example of this occurs when moving the nodes of a plane. In this case, the plane equation is already defined so, irrespective of the 3D cursor plane, such nodes will only ever move within the plane of the parent object.

Similarly, window and door entities inside a wall are also locked to movement within the plane of their parent object. In addition, no nodes are allowed to move outside the defining 3D polygon of the parent.

There are other more complex situations, such as when moving the nodes of an extruded object. In this case, each node will only move along the line of the extrusion vector. The full implementation of both the 3D cursor system and object snapping is discussed in Section B-7.

Relationship Mapping

Defining automatic geometric relationships between building elements can reduce data entry time and substantially increase model editability. It basically means deriving the geometry of one element from the geometry of another, and storing the rules used. If these rules are edited at a later time, or the parent element moved, the geometry of the child is automatically updated. If implemented correctly, for example, moving part of the floor plan or adjusting its height should automatically update the walls and ceiling of a space. Even changing the height of the first stair riser should update all others.

These relationships also extend to child objects such as windows and doors. Any movement of a parent object must automatically update the position of child objects. This is a simple matter during rotations and translations, however, resizing and rescaling must also be accommodated. The full implementation of a relationship mapping system is discussed in Section B-10.

Nodal Manipulation

The manipulation of the individual nodes that make up an object is an important part of any interface as it allows for increased flexibility in what geometry can be modelled. The user can take advantage of the efficiencies offered by a parametric base, but be able to add, delete or move individual nodes to form a much more complex element.

Similarly, groups of elements can be stretched by simply selecting a group of nodes and moving them together. This is very similar to the techniques used in most 2D vector graphics packages. It also greatly improves the editability of the model once created. The selection and manipulation of object nodes is discussed in Section B-13.

Graphical Display of Inputs and Outputs.

The very nature of the architectural design process is visual. This is especially true of the early stages of design where the building form itself is still being established. The ability to visualise a geometric model in three dimensions is therefore considered very important.

There are three elements to working in 3D:

• the ability to manipulate and change views quickly and interactively
• the ability to generate geometry in perspective
• the ability to analysis results spatially.

Establishing a 3D Viewpoint

The purpose of three dimensional representation of building models varies with different applications. In CAD programs, three dimensional and perspective projections are used primarily for the verification of model geometry and for presentation images. This requires two systems of viewpoint generation, a relatively simple method for selecting variable angle isometric and axonometric views and a more precise method of selecting perspective viewing positions and a view vector. Given options for cutting planes and camera lens settings, establishing the right perspective view of a model can be an involved and iterative process.

Rendering and visualisation programs have similar requirements to CAD. Geometric modelling requires the ability to quickly and interactively change viewpoint, whilst generating a rendered image usually requires the precise specification of camera position and lens settings.

In a conceptual design tool, precision is not as important as interactivity. Ideally, the process of selecting a three dimensional view should be as simple as examining an object in the hand, twisting and turning it at will. This assists the designer build up a spatial understanding of their design. Similarly, switching from perspective and orthographic projection should also be a simple and instantaneous process, especially when creating or positioning objects.

The notion of interactive view manipulation has been tackled by a number of virtual reality engines, emerging as plug-ins to well known web browsers. Whilst these are based on the premise that the user will want to travel around within a model, most provide an Examine mode to quickly rotate around and examine the model as described above. The implementation of this mode on many of these tools highlights two problems: What origin point and what axis to rotate around.

A significant amount of testing has been done in this work to develop a technique of model examination that is both simple and intuitive. This involved a number of methods of selecting the focus point and setting the sensitivity of the rotation angle to mouse movement.

The result is a system in which the focus point is always set to the centre of the view grid. This ensures that the model is always within view at all rotation angles. In addition, separating horizontal and vertical mouse movements into azimuth and altitude angles overcomes the problem of a changing ‘up’ vector. This is always maintained as the Z axis even though the view moves up and down. This way control is always retained and it is relatively easy to reset the view.

View rotation is simply a matter of dragging with the right mouse button in the model canvas. This can be done even whilst adding and dragging a node, simple press the right mouse whilst continuing to hold down the left. The perspective distortion is set using the view distance, the distance between the virtual eye and centre of the model. As discussed in Part B, this can be set interactively using the +/- keys while pressing the Control key.

Working in 3D

As discussed in Section A-3.4.1, the use of a fully 3D cursor system, together with snapping and relationship mapping between objects produces an environment in which modelling in 3D is relatively simple and intuitive.

Displaying Calculations in 3D.

Most existing analysis tools provide very little visual feedback during calculations. This means that the process being undertaken is essentially hidden from the user, who has to trust in the fact that what is being modelled is correct. Mistakes in modelling that are not immediately visually apparent must be determined from a close examination of any output.

Whilst the majority of calculations are not inherently visual, there are techniques that can be used to make them more so. For example, when using sampling or ray-tracing techniques, it is a simple matter to display each point or ray as it is generated and tested. This acts to provide an indication of how the calculation is progressing as well as allowing the user to identify possible problems with the model by observing anomalies in the display. Such techniques have been implemented in this application during surface area, volume, daylighting and acoustic calculations.

Multi-level Inputs

In the context of the workshop described previously, the desire for multi-level user input referred to the desire for users to specify almost a proficiency level, using which the software tool could adjust the requirement for detailed input for a specific calculation. Essentially novice and expert modes.

In a broader context, a more appropriate implementation of this is to structure calculations around a full set of basic assumptions and default values, any of which the user can change at any time. Inexperienced users, or those requiring a quick result, need only specify whatever level of information they have at the time.

This touches on two important and related areas, the potential for progressive data input and process modelling.

Progressive Data Input

Progressive data input refers to the ability to enter only a small sub-set of data required to model a particular process and generate results almost immediately. As the design is gradually resolved, more detailed information is added to the model, making the results progressively more accurate. This makes the process of modelling far more responsive.

There are, of course, issues with the validity of results based on default values, however, the same limitations are true of simplified manual and rule-of-thumb methods. These are well understood and accounted for by most practitioners. Where accurate results are more critical, more information is provided. This allows the designer to control both the effort and accuracy required for a result, not the application developer.

Given the detailed input requirements of most modelling systems, any information shortfall must be filled somehow by default or inferred values. There are many ways to derive or collect this data.

Using an Intelligent User Interface

A great deal of information can be inferred from the context in which an object is created or from the group of actions that may have preceded it. Similarly, it is possible to provide a number of alternate means of invoking the same action, each with different consequences for associated data. This way the user need only select a different icon or menu item in order to enter the same initial values, but with a completely different set of defaults. For the advanced user, the ability to change the default inferences may be appropriate.

Maintaining an Architectural Knowledge Base

Traditional CAD applications concentrate on the drawing process rather than modeling, the lines that define an element only provide clues as to its function. In a modelling tool however, knowledge as to the function of an element is essential. Such information facilitates the scheduling and cost tracking of components, automatic generation of building characteristics (percentage north glazing, floor area or lighting energy per metre squared), and allows analysis engines to react differently to different elements. This means that such an application must have some internal representation that a particular object is a floor, for example, as well as what it might mean to be a floor.

Whilst there are some fundamental elements common to most architectural designs, the correct assignment of elements must be left to the building designers ultimate discretion. It is a relatively simple matter to define wall, floor, ceiling, roof, window or door elements. However, how does one classify an external shading device, or a series of columns.

Automatic Data Gathering

Given a defined geometry, information such as the surface area of walls, the volume of spaces and areas of intersection are relatively simple to extract. Thus, if the 3D modelling interface can be made sufficiently simple and intuitive, inputting the geometry of a building can obviate a significant amount of numeric data input as this can be derived directly by the application when required.

Process Modelling

The convenors of the two workshops described in Section 3.2 highlighted what they called the ’schizophrenia of developers’. On the one hand, there was an almost universal desire for all modelling to be based on fundamental physical processes at the most detailed level. On the other, there was a concession that some simpler models requiring much less user input and computation time were sometimes just as valid.

This questions the necessity for the results of design tools to be perfectly accurate. Manual methods of estimating performance, such as those in the CIBSE Guide, contain quite a few highly simplified models that are known not to correlate well with actual physical processes. However, experience suggests that they usually result in a good design response.

David Bartholemew describes this as ‘validity defined as usefulness’ [Bartholemew 1997] and suggests that it is not actually necessary for a model to capture natural processes particularly well in order to be useful. What is important is the practical application and relevance of its output.

Absolute and Relative Accuracy

It is argued here that there needs to be at least two levels of modelling and analysis. The first, early in the process, to provide interactive design feedback. For quickly testing the viability of an idea, the comparison of multiple options and even the preliminary estimation of element sizes.

The absolute accuracy requirement at this level is quite low. What is more important is its relative accuracy, being able to immediately assess changes resulting from a particular set of design decisions compared to the original condition. This acts to guide the decisions-making process in the right direction. At this stage, all that is required is an indication that a problem may exist, its absolute magnitude can be the subject of second level analysis if it becomes more important.

The second level of analysis needs a more detailed and comprehensive model and is more likely to be based on fundamental physical processes. In this case absolute accuracy is important. This is the level of existing design tools and some interoperability with the conceptual design tool is essential.

Use of Components and Library Data

The exchange of electronic data is becoming more important in all industries. The ability to exchange ideas and design information with other architects and consultants is an essential part of any practice. To facilitate this, it must be possible to encapsulate all of the data required to describe a building model so that the receiver is not disadvantaged by not having the same weather files or material libraries.

As a result, an attempt has been made in this work to include all of the information that describes the model and its analysis in the one file. The file itself is structured into a series of ‘chunks’. Each chunk has a small header and contains information relating to either building geometry, material specification, calculation results or climatic data.

This internal structure has two advantages. The first is that other applications, which may not require the entire data set in a file, can search through and respond only to the chunks they actually want. Secondly, version information can be included in each header in order to maintain some backward compatibility between product releases.

A preliminary analysis of compression algorithms showed that the structure of data within a file is an important factor determining overall compression ratios. Many encoding routines work on large files by breaking them up into discrete segments, usually with a size based on a power of 2. Similarly pattern matching is also done in small segments.

In order to optimise the compression ratios within the data files produced by this application, care has been taken to ensure that the data structure of each object is byte aligned on an even byte number and that, when stored, these are each sized to a power of two. This was allowed for in the design of each node, object, zone and ray data structure to ensure that there was minimal increase in file size. The result is that compression ratios for the model files used in this application average between 8:1 and 12:1 using LZW and Hufman type encoding.

Material Libraries

The design of the material library within the application makes provision for specific manufacturer data to be included. At the simplest level this means space for a manufacturer’s name and an order number.

At a more complex level, it allows manufacturers to specify and lock the dimensions of their own library objects. This means that, for example, a library containing a range of aluminium windows could be provided that has accurate performance specifications for each element based on actual glass/frame ratios and production costs.

With more work, it is hoped that this type of consideration will induce manufacturers to provide material libraries of their own product ranges and maintain up-to-date versions on web sites or mailing lists.

Interoperability with Other Tools.

At the most simple level, interoperability means providing some basic interface with another application. The most usual form of this is the output of data in a file format that can be read by the other application. A more sophisticated system may use inter-process communication protocols, such as IPC on a unix system or Dynamic Data Exchange (DDE) in Windows, to transfer the data directly.

Interoperability with CAD systems can be achieved by supporting the DXF file format, a defacto standard amongst such applications. Similarly, this allows communication with a range of rendering and visualisation tools. Support for more task-specific tools such as RADIANCE may require the output of model data in its own native format. In these cases, greater interoperability would mean providing a means of controlling parameters important to the generation of the data file.

A further level of interoperability would be to recognise features in other applications which can be used to create efficiencies. One example of this is in the generation of RADIANCE data. As this tool uses radiosity techniques to calculate the distribution of light on each surface and then iteratively solve for inter-reflection, a wide variation in object size can lead to inaccuracies and increased calculation times. As a result, a means of breaking up large surfaces (such as floors and ceilings) into a number of much smaller ones can greatly increase the accuracy of generated results, as shown in Figure 3.2.

Figure 3.2 - Interpolation of geometry into output more efficient for use by other applications such as RADIANCE and CFD tools.

A further example of this is the interface with computational fluid dynamics applications. The complexity and efficiency of the form-fitted grid is integral to the amount of time required to converge on a fluid flow solution. If the output data is already aligned to specific grid points, the task of creating the optimum grid can be automatically carried out by the CFD application itself.

The required data is also significantly different in format from the geometric data used to model the building. Using Cartesian coordinates, angled planes must also be reduced to a complex set of axially aligned planes. Including this level of intelligence within a tool can make a relatively simple task out of jobs that would not normally be considered. Figure 3.3 shows an automatically generated CFD model of an entire inner city development being used for wind studies.

Figure 3.3 - Inner city development model, automatically orthogonalised by conceptual design
tool to create data more appropriate for an efficient CFD solution.

The Implementation

The following is a summary of the algorithms selected for use in the performance analysis of the geometric model. As these are mostly implementations of widely published algorithms, only details specific to their use within this work is discussed.

Automatic data gathering.

Using only a simple set of geometric analysis algorithms, it is possible to automatically derive a significant amount of data that the user would normally have to enter manually. This automatic generation of data is an important aspect of new generation design tools, making them much more efficient and easier to use.

Surface Areas

The surface area of a plane is automatically calculated each time its geometry changes as a result of rotation, scaling or the movement of any of its nodes. The algorithm to calculate this needed to be able to handle polygons of any degree of complexity, both concave and convex, with any number of vertexes.

The method chosen is based on the fact that the length of the cross-product of any two vectors is equivalent to the area of a parallelogram formed by two parallel and equal vectors. Thus, the surface area of a triangle can be calculated as half of the cross-product of any two of its sides.

Figure 4.1 - Diagram showing the length of the cross-product of a triangle as
equal to the area of an equivalent parallelogram.

The surface area of a complex polygon is therefore calculated by summing the cross-product of each triangle formed by subsequent vertices. In a concave polygon, the direction of the cross-product of each consecutive triangle will change.

Figure 4.2 - Diagram showing the surface calculation method.

By summing each cross-product as a vector, the overall surface area of any polygon is given by halving the absolute value of the total, taking into account the areas of child windows, doors, panels and voids.

Volumes

Volume calculations had to be very resilient and able to handle zones that are not fully enclosed or have overlapping surfaces. As a result, a sampling algorithm was chosen.

Using this method, the extent of the zone is first calculated and then grided against one of the primary axis. The resolution and axis of this grid can be specified for each zone. A pseudo-random ray is then generated somewhere within each grid square and tested for intersection with each planar element belonging to that zone.

The distance between the two extreme intersection points, multiplied by the area of the grid square, represents the volume contributed by that square. If less that two objects are intersected, the volume contribution is zero. The sum of all contributions gives the volume of the zone.

The accuracy of this algorithm can be set by the user and is very quick to calculate. However, it does not handle very thin concave polyhedrons well so the grid axis must be manually chosen to compensate for this. Each volumetric ray is displayed as it is generated. This way, real-time visual feedback can be used to determine if an axis is inappropriate for a particular zone and a more appropriate one selected. For almost all building applications, however, the default Z axis is usually the most applicable.

Figure 4.3 - Pseudo-random volumetric calculation algorithm in progress.

Intersection Points

The intersection point between a line and plane is a very simple calculation. Determining if this point falls inside or outside a complex 3D polygon is slightly more difficult.

It is possible to test if a point is internal to a polygon by summing the angles created from that point and each subsequent set of two vertexes. The direction of the normal to each equivalent triangle is calculated and determines if the angle is added or subtracted. If the absolute value of this summation is equal to 2 PI, then the point is internal. If it is zero, the point is external.

Figure 4.4 - Diagram showing angle-based method of determining whether a point is internal or external to a polygon.

Adjacency

Adjacency means determining the intersection area between two planes. This must be determined, for example, when two zones butt up against each other. Whilst it is possible to work out the exact profile of a complex intersection area, a sampling algorithm was used as it proved more resilient and is a simple method capable of handling any particular configuration and level of complexity. It also provided a useful level of visual feedback during the calculation to allow its validity to be checked in real-time.

Using this method, a small amount of pre-processing is performed to determine the number of objects that share the same plane equation. A number of points are then generated in pseudo-random positions within a grid over the entire extent of the test surface. Any points internal to two or more planes contribute the surface area of their grid square to the adjacent portion. The sum of all contributions gives the total area of intersection.

Figure 4.5 - Diagram showing adjacency calculation in progress.

Adjacent FLOOR-FLOOR or CEILING-CEILING objects are not tested as the result is meaningless and indicates intersecting zones or an error in zone layout. In most buildings, floors and ceilings share the same plane so all ceilings would have to be tested against each other, significantly increasing calculation time for no gain. For zones above and below, ceilings should normally be adjacent to the floor above and visa-versa.

Accurate knowledge of adjacency is important as geometric elements can be defined as having one material when exposed to the outside, and another when adjacent to another building zone. For example, a wall may be cavity brick when exposed or plaster-coated single brick when adjacent. This allows zones to be interactively dragged around and repositioned in their entirety without having to adjust walls divisions. In addition, this is vital if adjacent walls from different zone are not to be double-counted during cost analysis and thermal mass calculations.

Shadow Analysis

Shadows are generated from geometric objects by projecting vertices from a virtual sun onto a receiver plane. The azimuth and altitude of the virtual sun is calculated using formulae first proposed by Spencer [1965]. Values for solar declination and the equation of time are determined using formulae proposed by Carruthers, et al [1990], as shown in the following pseudo-code:

// Solar declination as per Carruthers et al.

t = 2 * M_PI * ((iJulianDate - 1) / 365.0);

fDeclination = (0.322003

		 - 22.9711 * cos(t)

		 - 0.357898 * cos(2*t)

		 - 0.14398 * cos(3*t)

		 + 3.94638 * sin(t)

		 + 0.019334 * sin(2*t)

		 + 0.05928 * sin(3*t)

		 );

// Convert degrees to radians.

fDeclination = (fDeclination / 180.0) * M_PI;

// Equation of time as per Carruthers et al.

t = ((279.134 + 0.985647 * iJulianDate) * M_PI) / 180.0;

fEquation = (5.0323

		 - 100.976 * sin(t)

		 + 595.275 * sin(2*t)

		 + 3.6858 * sin(3*t)

		 - 12.47 * sin(4*t)

		 - 430.847 * cos(t)

		 + 12.5024 * cos(2*t)

		 + 18.25 * cos(3*t)

		 );

// Convert seconds to hours.
fEquation = fEquation / 3600.00;

// Difference (in minutes) from reference longitude.
fDifference  = (RAD2DEG(fLongitude - fTimeZone) * 4) / 60.0;

Sunrise and Sunset

Sunrise and sunset times are calculated symmetrically about solar noon using the method outlined by Spencer [1965]. In this method the actual times of sunrise and sunset are considered to occur when the very top edge of the sun is parallel with the horizon. As a result, 0°50’ is added to sunrise and sunset angles to allow for the radius of the sun. Sunrise and sunset times are displayed in local time.

Horizontal and Vertical Shadow Angles (HSA & VSA)

Horizontal and vertical shadow angles are calculated for individual planes using the method proposed by Spencer [1965]. This same method appears in several other publications [Phillips 1983, Szokolay 1996], as shown in the following pseudo-code:

HSA = fSolarAzimuth - fNormalAzimuth;

top = sin(fSolarAltitude - fNormalAltitude);

bot = cos(fSolarAltitude - fNormalAltitude) * cos(HSA);

VSA = atan(top/bot);

Viewing From Sun

The design of shading devices and the analysis of sun penetration can benefit significantly from an appreciation of the model from the sun’s perspective. As the sun’s rays are very close to parallel when they hit the Earth, this involves generating an orthographic view of the model from the same azimuth and altitude as the sun. This view can be extremely useful when manually creating or editing solar shades or simply determining overshadowing.

Figure 4.6 - East facing shading system designed using view from sun position.
Note the change in shading depth from north to south.

Overshadowing

Overshadowing analysis is simply a matter of projecting shadows onto the ground or on specific planes tagged as shaded surfaces. Using tagged planes allows both overshadowing and internal sun penetration to be displayed. Figure 4.7 shows two overshadowing examples whilst Figure 4.8 shows an instance of both sun penetration and a reflective light shelf.

Figure 4.7 - Examples of overshadowing projection on the ground and on a contoured surface.

Figure 4.8 - An instance of both internal sun penetration and a reflective lightshelf.

Isolating Individual Shadows

During overshadowing analysis, it is often necessary to isolate or highlight the shadows of one or more buildings from amongst many. This can be done in this application by simply varying the transparency of key buildings. In the following example, the background buildings have been made 50% transparent whilst one was made 25% and another fully opaque.

Figure 4.9 - The isolation of buildings for either total or additional overshadowing.

The second image in Figure 4.9 shows only the additional overshadowing created by the two buildings. This is achieved by holding down the Shift key when selecting shadow display, which simply reverses the sort order of shadow polygons. This type of analysis is essential for inner-city councils when assessing new building proposals.

Shadow Profiles

Another important aspect of overshadowing is the shadow profile. This is simple the outline of shadow produced by a building over time. This is displayed as a sequence of shadow diagrams which can be overlaid to form a single image, or output to sequential files to form an animated GIF or AVI movie file.

Figure 4.10 - A shadow profile displayed over a specified time range

Stereographic Analysis

Sun path diagrams are an extremely useful tool for both the siting of buildings and the design of shading devices. In the one diagram, overshadowing for an entire year can be displayed. These are constructed by projecting the hemispherical sky dome onto a flat diagram, usually circular. A number of methods are available for translating solar altitude into a diagrammatic radius. Stereographic projection is the most widely used method and has been adopted in this application as it provides the greatest accuracy at low sun angles. A comprehensive analysis of alternate methods can be found in the Environmental Science Handbook [Szokolay 1980].

The stereographic diagram used here, as shown in Figure 4.11, is constructed directly from solar geometry. The path of the sun through the sky dome is based on its actual altitude and azimuth calculated at each hour for the selected location. Thus hour lines show the characteristic figure 8 of the analemma. The two halves of the year are differentiated using solid and dotted lines.

Figure 4.11 - Stereographic diagram showing sun paths and overshadowing patches.

The projection of model geometry onto the stereographic diagram involves tracing each line as a number of discrete segments. This describes each straight line as a curve when projected onto the sky dome. This is more accurate than a number of simplified methods in common use that only project object vertices [Szokolay 1980, Phillips 1983].

The displayed diagram is only valid for a single focus point. If a planar object is selected, the diagram is generated from its geometric centre and a warning stating this is displayed.

The generation of shade patches this way is still quite quick, allowing the diagram itself to be interactive. Thus, moving or dragging the focus point within the model automatically updates the display. This level of interactivity means that a full overshadowing analysis at numerous points around a site can be carried out in minutes once a model has been built.

Solar Exposure

Once the position of the sun and incidence angles have been determined, the solar exposure of any surface can be calculated. This is done as either instantaneous irradiance or integrated irradiation over an entire day, month or year. In order to determine the amount of incident radiation, the following parameters are required:

• Global, beam and diffuse solar irradiance
• Percentage shading or overshadowing
• Angular dependant reflectivity, if applicable

Global, Beam and Diffuse Irradiance

Values for global and diffuse solar irradiance on a horizontal surface are calculated in two alternate ways depending on their use. When used as part of the thermal analysis of the model, hourly values for each day of the year are read directly from the location data file. These values are based on recorded meteorological weather data and are linked directly to cloud cover, sol-air temperature and absolute humidity.

When explicitly calculating instantaneous solar exposure for comparative analysis, recorded irradiation data can be misleading as values are used in isolation, with no information as to other external conditions such as cloud cover. In order to average out spurious fluctuations in these cases, values for global and diffuse irradiation are derived from average daily irradiation figures, as displayed in the monthly climate summary. The method used to derive instantaneous global and diffuse values this way is given by Szokolay [1987].

Once these two values have been read from the location file or derived from averaged monthly data, the amount of beam irradiance on a surface normal to the sun can be calculated using a method also described by Szokolay [1987].

To accurately quantity solar exposure, the effects of direct, diffuse and reflected irradiance must be considered.

Direct Irradiance
The direct component is that irradiance which results from direct exposure to the sun. This depends on the angle of incidence of the direct irradiance as well as the percentage of the surface currently in shade. The method of determining percentage shading is described in Section A-4.4.2.

Diffuse Sky Irradiance
Diffuse irradiance refers to that component of the total that arrives from all angles over the entire sky dome. This is dependant on the tilt angle of the surface. Obviously horizontal surfaces are exposed to the entire sky dome whilst vertical surfaces are exposed to only half. This is a simple linear relationship with tilt angle.

The effects of geometric obstruction and overshadowing on the diffuse component is difficult to determine geometrically. An accurate method would require determining the average percentage of the sky dome visible from all points over the entire surface. The calculation of this value even for a single point is quite computationally intensive.

As a result it is assumed in this application that, whilst the diffuse component will have some effect on solar collection and sol-air temperature, this effect is not deterministically significant. Thus no consideration of overshadowing is applied to the diffuse component.

This can lead to some anomalies within the model. For example, a floor element whose surface normal faces upwards would receive full diffuse radiation from the entire sky dome, even though almost entirely obscured. As the small amount of diffuse radiation that may impact on the floor is picked up by the windows that it must first pass through, objects defined as a FLOOR are not considered in diffuse radiation calculations. Similarly, any portion of a planar object that is adjacent to another zone is also not considered, regardless of the orientation of its surface normal.

The application itself automatically orients the surface normals of objects created within its own 3D interface. Thus, correcting orientation is only a consideration when using imported geometry from DXF files and CAD models.

Ground Reflected Irradiance
Ground reflected irradiance is simply that component of the total that is reflected off the ground plane. This is dependant on tilt angle in the opposite way to diffuse sky irradiance. In the application, no consideration for geometric obstruction is applied to this component either.

A default ground reflectivity of 0.2 is assumed unless explicitly set in the Preferences dialog box.

Shading Factor

The shading factor for a surface is calculated geometrically and is given as the percentage of that surface currently shaded from direct sun. This is calculated by generating a series of pseudo-random points distributed over the surface. Geometric rays are then traced from each point towards the sun.

If an opaque object is intersected at any point, that ray returns a value of one (1.0). If only transparent objects are intersected, the ray returns the cumulative effect of each object’s transparency (<1.0). If no objects are intersected, the ray returns a value of zero (0.0). The shading percentage is simply the sum of the return values of each ray divided by the number of rays generated.

Figure 4.12 - An example of shading factor calculations being used to determine the
transmission characteristics of angled shading mesh at different times of the year.

In order to optimise calculation time, the distribution density of rays is determined from the entire model extents. This means that smaller objects generate less rays. This reduction in accuracy is in keeping with the less significant effects of smaller objects. The current distribution density can be set manually, and is actually displayed as white dots on the surface whenever the instantaneous exposure on a single object is calculated.

Solar Exposure Graphs

Solar exposure values can be graphed for any time of the day at any date. This information can be displayed as either hourly values throughout the day or as an average hourly distribution for each month of the year. These graphs show hourly values for total available irradiation, actual incident irradiation, any portion reflected off tagged surfaces and the percentage in shade for the selected object. Average hourly distributions can be shown for each of the above to understand when peak direct solar gains occur during the entire year.

Figure 4.13 - Graphs of daily and annual solar exposure for a selected surface.

Angular-Dependant Reflectivity

If the object receiving the radiation is defined as a transparent WINDOW, an additional angular-dependant transmission factor is applied to the beam component to account for changing reflectivity with incidence angle. This is based on Fresnel’s expression [Lynes 1968]:

reflected = 0.5 * ((sin²(i-r)/sin²(i+r)) + (tan²(i-r)/tan²(i+r)));

where r  = Internal refraction angle (r = sin(i)/refractiveIndex).

and   i = Incidence angle.

The refractive index is a basic material parameter that can be set for each glass type within the application. A default refractive index of 1.52 (based on 6mm clear float) is used if this value is not set.

An increased reflected component acts to reduce the amount of radiation transmitted through a transparent element. Thus angular dependant reflectivity of glass is an important consideration when calculating direct solar gain though windows and the effects of coverings on a solar collector. This effect has been incorporated into both the shadow and reflection routines. This can be observed as changes in sun patch brightness at sharper sun angles.

Shading Device Design

A simple methodology for the automatic generation of optimised shading devices is implemented in the application. Given a window and a limiting set of dates and times, it is possible to determine the exact geometry of an optimal shading device. This is simply a matter of determining the plane equation of each shade and then projecting the path of the sun onto it. This must be done for each vertex of the selected window. For horizontal shades, the effects of the analemma must be considered at each side if the shade is to work over a range of dates.

Once all of the vertexes have been projected at the selected date and each limiting time, a modified convex hull algorithm is used to determine the final shape of the required shade. A modification to this algorithm was required as the convex hull shape is not truly optimum. In the case of a rectangular window facing roughly north, the edges of a horizontal shade are actually concave at some points, as shown in Figure 4.14.

Figure 4.14 - Detail showing sun path lines from which the modified convex hull algorithm generates the required profile.

This same method can be used to generate shades at any angle and with both horizontal and vertical elements. Optimised solar pergolas are also possible as the altitude of the sun when normal to the window can be determined at both the current date (used for the cut-off angle) and in mid-winter (used for maximum penetration).

Figure 4.15 - Examples of automatically generated shading devices.

Solar Profiles

There will be times when such a parametric shade is not suitable for a particular application. To allow for this, sun paths can be projected onto any surface from any point. This creates a solar profile line which is clipped to the extent of the selected surfaces. Figure 4.16 shows an example of the use of such profile lines to design a complex shading hood that exactly shades the bottom corners of a rectangular window.

Figure 4.16 - A complex shading hood optimised using solar profiles
projected from each corner of the window.

Solar Extrusions

When considering overshadowing and solar access rights, it is often necessary to limit the height of a building based on a particular solar time. This can be easily achieved by extruding a plane towards the sun at some set of limiting dates and times. These planes can then be used to automatically cut the building envelope to ensure overshadowing does not occur.

Figure 4.17 - The use of projected solar access planes to control the overshadowing
effects of new buildings.

Daylight Factor

An estimate of daylighting levels and glare within a building is fundamental to fenestration design. Whilst the accurate calculation of such levels is quite computationally intensive, a manual method based on the British Research Station (BRS) Daylight Factor Protractors is widely used as a first approximation.

The Daylight Factor at a point within an enclosure is a function of three components, the sky component, externally reflected component and internally reflected component [Longmore 1968]. The calculation of Daylight Factor in this application uses a geometric method for determining the sky and externally reflected components and the standard BRS formulae for the internally reflected component.

Geometric Method

At each measurement point a series of geometric rays are generated. These rays are evenly distributed over the surface of an imaginary dome using an iterative geodesic triangulation technique. The altitude of each ray and the equation defining the type of sky (CIE Uniform or Standard Overcast sky) is used to determine the reference return value for that ray.

If the ray does not intersect any object and has a positive altitude, it returns the full reference value to the sky component. If one or more transparent objects are intersected, the return value is modified by the cumulative effect of each object’s transparency and is also added to the sky component.

If an external opaque object is intersected, the ray is terminated and the return value further modified by the external surface reflectivity of that object and added to the externally reflected component. The altitude of each such ray is also added to a running average maximum elevation of external obstructions for each window. This is for later use in the determination of the internally reflected component. If an internal object is intersected, the internal surface reflectivity of that object is added to a running average values.

Figure 4.19 shows a daylight factor calculation in progress. As each ray is displayed as it is traced, this visual feedback can be used to check the validity of the model, making sure rays travel through windows and do not ‘leak’ out of opaque parts of the enclosure.

Figure 4.18 - A partially complete calculation of Daylight Factor at points within an atrium model.

Once all rays have been traced, the ratio of the total sum of all possible return values versus actual return values becomes the sky and reflected components.

Internally Reflected Component

The internally reflected component is calculated using the standard BRS formulae. This formulae is based on average surface reflectivity’s at different heights and returns a room averaged value that is independent of actual position. This is the same formulae used in published manual methods.

The parameters used in this formulae are collected during the geometric calculation, being the average reflectivity of surfaces above and below point height as well as the average elevation and reflectivity of external obstructions.

Translation Into Illumination

If the daylight factor at a particular point and the design sky illuminance for the current location is known, then a representative value for illumination for a uniform or CIE overcast sky can be calculated. This is simply the product of the two, given in Lux.

The resulting value is neither a minimum nor an average, but a representative design value of useful interior daylight available over approximately 85% of the working year (taken from 9:00am to 5:30pm).

This relationship is useful as it can be reversed to actually determine a daylight factor. If the required illumination for a particular task is known, then the daylight factor required to provide that level 85% of the time is simply given by:

DaylightFactor = DesignSkyIlluminance / RequiredIllumination;

Artificial Lighting

Artificial lighting levels are calculated using the point-by-point method. This is a geometric algorithm in which the contribution of each light source is simple summed at points within the enclosure. The luminance distribution of each light source is used to modify contributed levels based on the off-axis angle of each point. The inverse square law is then used to account for geometric spreading.

This is a simple method for determining direct light levels from both regularly and irregularly spaced luminaires. It is a useful tool for ensuring adequate minimum light levels and is mainly intended for use in this application as a comparative measure. This method takes no account of diffuse light or inter-reflection between illuminated surfaces.

The lumen method of artificial lighting design is not implemented as it is considered to promote an inefficient distribution of luminaires where localised task lighting may be more suitable. The point-by-point method can be used in place of the lumen method to design and analyse regular arrays as well as localised lighting.

RADIANCE Output

For a more accurate analysis of both natural and artificial light levels, the geometric model can be converted into a RADIANCE model. RADIANCE is a public domain lighting simulation tool produced by Greg Ward at Lawrence Berkeley Laboratories. It is widely used for lighting design and has been extensively validated. Images are generated using the fundamental physical properties of light transfer and inter-reflection based on a hybrid radiosity/raytracing technique.

The conversion routines within the application are quite sophisticated, making use of primitives native to the package and even generating a RADIANCE Instruction File (RIF) to control program execution. RADIANCE is freely available for Unix workstations and for DOS machines as part of the ADELINE package.

Figure 4.19 - An example of a geometric model and corresponding RADIANCE image.

Thermal Performance

To calculate internal temperatures and monthly heating and cooling loads, the Admittance Method is used as described in the CIBSE GUIDE Volume A, Design Data [1986]. Some methods of accounting for solar gain are taken from Szokolay [1987], as described in Section A-4.4.1.

This method is based on a steady state analysis and simulates the dynamic response of a building by calculating cyclical variation about a 24 hour mean temperature. Steady state conditions are used to determine the relationships between energy flows, the thermal characteristics of each space and air temperatures. The thermal response of the building fabric is then used to determine the dynamic response of each space.

This method was selected as it strikes a balance between accuracy and simplicity, being quick to calculate whilst providing quite detailed performance feedback.

It should only be considered as a first level approximation of thermal performance. As such, the absolute accuracy of the results are not that important. However, the relative accuracy of the results are very good, allowing for the detailed comparison of options to guide the design as it is being developed. Interoperability with more sophisticated applications that use the response factor method are currently being developed.

Internal Temperatures

As a first level approximation, the admittance method is an excellent tool. It allows the calculation of internal temperatures with full consideration of:

• The effects of thermal mass,
• The effects of internal radiant temperatures,
• The effects of air infiltration and static ventilation rates,
• The effects of occupancy and equipment gains,
• Direct and indirect solar gain.

From this calculation, a 24 hour temperature profile of each zone in a building can be determined. The influence of adjacent zones at different temperatures is not well defined in the CIBSE Guide, except as an additional gain to be added to the space load. As the internal conditions of each zone in this application are undefined at the beginning of a calculation, and the order in which zone temperatures are calculated is arbitrary, it is not possible to properly consider inter-zonal flow during the first calculation. As the method is steady state based, pre-conditioning each zone by starting the calculation a number of days earlier is of no significant benefit. However, a single day’s preconditioning is always performed to obtain an approximate temperature record for the previous day. This assumes that no building element has a thermal lag greater than 24 hrs.

As a result, once the internal temperatures have been determined using the base CIBSE method, an additional set of iterations are performed using stored adjacency data to update inter-zonal heat flows. Each calculation is very simple and continues until the maximum change in each zone is less than 0.1 of a degree.

Figure 4.20 - An example of a 24 hr temperature profile of a house in summer
calculated using the Admittance Method, picking up the relatively high
internal temperatures of its roofspace.

Heating and Cooling Loads

Heating and cooling loads are calculated using internal air volumes and the difference between internal dry-bulb temperatures and the heating and cooling thermostat settings for each zone. This provides an estimation of the required sensible loads to maintain bulk air temperature within the specified range.

Figure 4.21 - An example heating and cooling load estimation
for the same house as in Figure 4.20.

It is important to note that these values are not to be considered as plant loads. They are simply sensible space loads. No allowance is made for latent gains due to occupancy or equipment. Estimated costs are calculated based on electrical utility rates and the specified system efficiency. They are displayed simply to provide a more meaningful value to the results and facilitate the approximate cost/benefit analysis of various options.

Passive Design Analysis

Heat load analysis assumes the building is air-conditioned. Whilst this is reasonable for commercial buildings, it is not so for many residential and industrial buildings. Some form of performance feedback for passively controlled buildings is also necessary in a conceptual design tool.

This is achieved by generating a cumulative frequency graph of internal zone temperatures showing the number of hours per year that the temperature was a particular value. The vertical scale thus represents the number of hours whilst the horizontal scale represents temperature. For example, the internal temperature of a particular zone may have been between 20-21°C for 1325 hours, whilst between 30-31° for only 86 hours.

Figure 4.22 – A cumulative frequency graph of internal temperatures for the
analysis of passively controlled buildings.

The white area in the centre background of the graph indicates the comfort band set for the currently selected zone. The current zone is highlighted in red with all other zones shown in yellow. The distribution of outside air temperatures is also displayed as a dotted blue line for direct comparison.

Using this graphical technique, it is possible to quickly appreciate the performance of unconditioned zones as well as zones conditioned only part of the time. If the graph is highest within the lightest gray area, then the zone is performing well, remaining within its specified comfort zone most of the time. If not, it either requires more heating or cooling, depending on which side of the comfort band.

Isolated Component Analysis

In order to properly devise a passive design strategy, some feedback as to the relative contribution of each component of the thermal stress on the building is required. This is provided in the form of an average hourly distribution of temperatures and gains for each month of the year, as shown in Figure 4.23.

Annual distribution of conduction gains

Annual distribution of ventilation gains

Annual distribution of direct solar gains

Figure 4.23 – A comparison of the annual distributions of fabric, ventilation and
solar gains in a passively controlled building.

These show the maximum and minimum effects of each of the following components throughout the year:

• Average hourly internal temperatures,
• Average hourly conduction gains,
• Average hourly direct solar gains,
• Average hourly solar excess gains as a result of radiation on opaque elements,
• Average hourly internal gains from appliances and equipment and
• Average hourly ventilation gains.

Statistical Acoustics

Having the geometric model of a building, within which the internal volume of each space is known as well as the surface area and material definition of each element, the calculation of statistical reverberation times becomes a trivial task. As a design tool, the room averaged reverberation time is still one of the first objective measures to be calculated, even by specialist acoustic consultants. In some cases, it is the only objective measure calculated.

Within the material database, the Sabine absorption coefficients for each octave between 63 Hz and 16 kHz are stored for each material as shown in Figure 4.24.

Figure 4.24 - An example of the sound absorption coefficients
stored for each material.

As part of the calculation, an analysis of the distribution of absorption coefficients is performed in order to determine the most appropriate method to use. There are three possible methods, the standard Sabine formula for fairly reverberant rooms, the Norris-Eyring formula for rooms in which high absorption that is similarly distributed on all surfaces, and the Millington-Sette equation for rooms with widely varying absorption. A more detailed explanation of the implementation and use of these formulae is given in Section B-23.

An example of the calculated RT results for a space is shown in Figure 4.25, together with the accompanying text output.

Figure 4.25 - A statistical reverberation time graph for an auditorium with a widely varying distribution of sound absorption.

//
//          CONCEPT Data file.
//          Reference: West Australian College of Advanced Education
//          Description: Dramatic Arts Auditorium
//          Model File: c:\Concept\examples\drama.zon
//          Date: Oct 23 03:13:04 1995
//       

Zone:          Main Hall       
Volume:          7298.279 m³
Surface          Area: 3223.010 m²

Optimum          RT (@500Hz for Speech): 1.06s

Optimum          RT (@500Hz for Music): 1.75s

Volume          per Seat: 3.649 m³

Minimum          (For Speech): 5.94 m³

Minimum          (For Music): 9.38 m³

Method:          Millington-Sette (Widely varying)

RT          63Hz: 	A 513.74	2.20s 

RT          125Hz: 	A 536.88	2.13s 

RT          250Hz: 	A 978.84	1.17s 

RT          500Hz: 	A 1350.37	0.76s 

RT          1kHz: 	A 1460.89	0.64s 

RT          2kHz: 	A 1588.58	0.59s 

RT          4kHz: 	A 1565.41	0.59s 

RT          8kHz: 	A 1545.61	0.60s 

RT          16kHz: 	A 1614.67	0.57s 

Acoustic Raytracing

The basis of the geometric approach to acoustic analysis used within this application is discussed in an early paper by the author, included here as Appendix A. To summarise, geometric acoustics can not be used to fully describe all aspects of the acoustic performance of an enclosure. However, given the very strong relationship between room geometry and acoustic behavior, and the fact that at the conceptual stage of design it is that very geometry that is being decided upon by the designer, any information as the acoustic ramifications of design decisions is considered invaluable.

As a result, a wide range of geometric acoustic analysis methods have been implemented to assist with enclosure design. These range from basic geometric analysis to the reverse integration of the impulse response to obtain decay times.

Sprayed Rays

The most basic form of geometric analysis is to simply generate a number of rays within an enclosure and trace their path. Making such a method quick to calculate and relatively interactive, allowing the user to drag the source point to different locations and having the rays automatically updated, starts to make this both an analysis and a design tool.

As a result, acoustic rays can be sprayed around an enclosure at any range of angles and to any reflection depth. Whenever the geometry of that enclosure changes in any way, the rays are automatically regenerated to reflect the changed conditions. This allows reflectors to be interactively positioned at the optimum angle and entire room plans to be optimised for an even first order distribution.

Figure 4.26 - Sprayed acoustic rays being used to indicate the distribution, relative level and relative delay of reflections off selected surfaces.

In addition to the geometric ray path, rays are displayed in colour. Rays with a delay relative to the direct sound of greater than 100ms are shown in red. A gradual change from yellow to red occurs for rays between 50ms and 100ms delay. Both relative delay and relative level can be displayed as text at the termination of each ray.

The Image Method

A slightly more complex analysis is to determine the possible reflection paths between a sound source and any number of receiving points. This involves an exhaustive test of all reflections off all surfaces within the model. This is done by generating phantom images of the source position by reflecting it about each plane in the model. The visibility of this image is then tested for each receiver point and the reflection sequence stored for each valid ray. For second order reflections the phantom images are then further reflected about each plane, and so on.

Calculation times using this method increase exponentially, being equivalent to the number of planes in the model to the power of the reflection depth. In a typical enclosure of around 100 planes, fourth or fifth order reflection are not unrealistic with a calculation time of between one and two hours.

As this method is exhaustive, it calculates all possible reflection paths up to the specified reflection depth. Once the geometric path of each ray has been calculated and stored, it is possible to determine its time delay and sound level relative to the direct sound at any frequency for which there is sound absorption data for the materials intersected along the way.

The relative delay is simply determined from the path difference between the ray and the direct sound arriving at the receiver. When calculating the speed of sound, a default internal temperature of 20°C and a relative humidity of 60% is assumed. To determine the relative sound level, the effects of geometric spreading, boundary absorption and the molecular absorption of air are applied. Sound levels are taken relative to the direct sound at each receiver point, so are shown as a negative decibel value.

Angle-Dependant Absorption Coefficients

A facility has been included in this work for the provision of angle-dependant absorption coefficients, as discussed in Appendix A. However, due to the relative difficulty obtaining these values for a wide range of commonly used building materials, the final application only allows the specification of single coefficients at each octave. Future work will reactivate this feature.

The Hybrid Image Method

Given the exponential increase in calculation time required by the Image Method, Vorlander [1989] proposed a hybrid method combining elements of random ray tracing with the image method. Using this technique, the point receiver is replaced by a sphere of finite radius. A number of random rays are then generated to a depth set by the user. Each segment of a ray is tested to see if it intersects any part of the sphere. If so, there is a significant chance that there exists another ray, reflecting off the same sequence of boundary objects, but passing exactly through the centre of the sphere. As both the current image position and reflection sequence is known, it is a simple matter to trace the new ray from the centre of the intersected sphere to the image point, testing to see if it is a true ray of not.

The benefits of this method are quite significant. Rays can be generated with as great a reflection depth as required without the corresponding exponential increase in calculation time. Once a ray has been generated, it can be tested against as many spherical receivers as can be specified within the enclosure. Thus the time spent calculating one ray may yield any number of true rays for any number of receiver points at any depth.

Ray Distribution Algorithms

Unlike the method of images, this algorithm samples only from the set of real rays within the enclosure and is by no means exhaustive. Therefore a number of algorithms for the generation of sampled rays have been developed. This means that rays can be selected at random, based on a fixed interval, restricted to a given range of angles or only off a specified plane. As a result it is necessary to keep in mind the bias that any of these choices place on the acoustic measures derived from a particular set of rays.

Experience with this algorithm has shown it to be a remarkably fast and effective means of overviewing the behaviour of an enclosure. Using a single calculation of perhaps 10000 rays at a depth of 32 reflections may take up to an hour, but will yield a sample impulse response for as many points of interest as are required.

As a quantitative tool, however, there can be no measure of the fraction of real images actually represented, even at the lowest orders of reflection. This can be overcome, to a degree, by increasing the radius of each sphere relative to the room volume in order to catch and test more rays.

Relating Room Geometry to Room Response

Once reflections have been calculated, the application simultaneously displays the echogram immediately beneath the geometric paths of each ray. Selecting a ray within the model with the mouse highlights both the ray and its echo. Similarly, selecting an echo also highlights the geometric ray.

This direct relationship between the room response and the geometry of each ray allows a significant amount of both objective and intuitive analysis. As well as simply getting a feel for how sound waves are being reflected within the enclosure, it is also possible to instantly detect spurious echoes and, by simply selecting it, instantly see the path it took, as shown in Figure 4.27. Rotating the 3D model using the right mouse button displays only the selected ray in 3D so a full appreciation of the surfaces it reflects off can be gained.

Similarly, holding the left mouse down whilst dragging it over the echogram continually highlights the closest echo. In order to quickly get a feel for the paths of high-level reflections, the mouse can be dragged near the top of the decay curve. To view the low level reflections, the mouse is dragged further down the curve. An example of this is included in the tutorial manual in Part C.

Figure 4.27 - Individual reflections can be selected from either the model or
the echogram. This allows the room response to be related directly to
room geometry.

Derived Information

There are a number of ways to display the information that can be derived from these reflected rays. The aim is to present results in a way that is not only meaningful, but can also serve as a basis for further interactive exploration.

Reflection Point Analysis

When one or more planes are selected, the reflection points of each ray with those planes can be displayed. This provides an indication of those areas of a surface that are actually contributing to the reflection of sound between the source and the receiver. It is possible to quickly cycle through the reflection points for each receiver position using the Ctrl+PageUp and Ctrl+PageDn keys.

In the unix version, the relative sound level of each reflection is indicated by shades of yellow. This allows the more significant reflections to be easily discerned. This is more difficult in the Windows version, so will be fully implemented in future work.

Figure 4.28 - Reflection points displayed on selected objects along with the distribution of incidence angles.

When reflection points are displayed, the graph immediately underneath displays the statistical distribution of reflection angle. This graph shows the percentage of rays reflecting off the surface in specified increments. Thus it is possible to determine if the greater proportion of rays are at normal or grazing incidence. This can help determine the most appropriate type of absorber to use on that surface.

Lateral Energy and Polar Diagrams

Another important aspect of the acoustic performance of any space is the angular distribution of reflections arriving at the receiver point. An even distribution throughout all angles provides a more intimate feeling of being surrounded by the sound. This information can be displayed as either the distribution of phantom images or as a polar diagram.

Once again, the user can interactively select images or polar rays with the mouse. Using the right mouse button, the exact path of the corresponding ray can also be examined.

Figure 4.29 - Reflections displayed as a distribution of phantom images or as a polar diagram.

Integrated Energy

Schroeder [1965] has shown that performing a reverse integration on the impulse response produces a smooth decay curve equivalent to the ensemble average of many such decays. It is therefore possible to display this integrated decay curve and to calculate lines-of-best-fit for use in the estimation of both reverberation time and early decay time, as shown in Figure 4.30. Such a graph can also be used to estimate how exponential the decay within an enclosure actually is.

Figure 4.30 - Reverse integration used to calculate RT₆₀ and EDT₁₀ using lines of best fit.
Note the use of range indicators to more accurately control the range over which the best-fit is applied.

It is possible to increase the accuracy of the lines-of-best-fit using two interactive cursor points to set the range over which it is calculated. These can be positioned such that deficiencies within the decay, due to a lack of higher order reflections, can be obviated.

Conclusion

The primary aim of this work was to produce something akin to an environmental design calculator. A tool capable of evaluating any aspect of a design, from the smallest of details to the overall environmental impact of materials and fabric. As such, this work represents a new approach to the integration of environmental engineering and building design, two pursuits that, until relatively recently, were synonymous.

There are other organisations attempting similar tasks. The IES package described in Section A-2 is a perfect example. What differentiates the work presented here, however, is its focus on conceptual design. This is the earliest most defining stage of design, where each decision has the most potential and the greatest effect in terms of environmental performance. It is hoped that this work goes some way towards filling the void in this area.

A tool such as that described in Part B is never actually finished. However, it is hoped that a solid foundation has been established as a base for its ongoing development.

It is fitting then that this thesis should conclude with a list of ideas for additional ‘features’ still left to implement:

• A joint research project has begun with CSIRO, using the 3D geometric engine developed here as a front end to their Cheetah and MIXC natural ventilation engines. This is seen as a prelude to more closely interfacing with other thermal analysis tools such as CAMEL, CheeNATH, TAS and DOE2.

• The inclusion of simple natural ventilation analysis. It is intended that this begin with the simple estimation of buoyancy and wind driven pressures at each aperture and then progress to bulk air flows in a multi-zone model.

• As the geometry is known, as well as internal surface temperatures at each hour of the day, the next version will use the same routines as for natural lighting calculations to determine radiant temperatures at each sensor point. This is simply a matter of storing the surface index and distance for each sprayed ray and updating radiant temperatures whenever surface temperatures change. With air flow data from MIXC, this would be used to map thermal comfort measures such as Predicted Mean Vote (PMV) and Percentage Dissatisfaction (PPD) throughout a space.

• This version includes the calculation of incident solar radiation on any surface. The next step is to more fully incorporate solar collector design and wind power generation using real recorded weather data and conversion efficiencies to determine energy availability. These could then be matched against load profiles (using appliance operational profiles) for a more accurate estimation of required storage capacity.

• More consideration of life cycle assessment and cost-benefit analysis using embodied energy, greenhouse gas emissions and even off-gassing as part of the equation. The geometric modelling system here is considered to form a useful link between the actual building design and the quantification of environmental and life cycle data.

• Calculating inter-zonal sound transmission and sound propagation within large indoor and outdoor spaces.

• The ability to define more detailed surface textures for output in VRML, RADIANCE and POV-Ray files.

Bibliography

Akin, O. 1978, ‘How do architects design?’, Artificial Intelligence and Pattern Recognition in Computer Aided Design, Latombe, J.C. (ed), North Holland.

Ashdown, I. 1994, Radiosity: A Programmer's Perspective, John Wiley & Sons, Inc, New York.

Bellchambers, H.E., Lambert, G.K. and Ruff, H.R. 1961, ‘Modern Aids to Lighting Design - Computer Techniques’, Trans Illum Eng Soc, Vol. 26, No. 3, pp 107-122.

Brittain, J. 1997, ‘The Cost of Oversized Plant’, BSRIA Guidance Note No. 12/97.

Bunn, R. 1995, ‘Tower of Babel of the Promised Land’, Building Services Journal, 1/95.

CIBSE, 1986, CIBSE Guide, Volume A - Design Data, The Chartered Institute of Building Services Engineers, Staples Printers, St Albans, England.

Carruthers, D.D., Roy, G.G. and Uloth, C.J. 1990, An Evaluation of Formulae for Solar Declination and the Equation of Time, The University of Western Australia School of Architecture, Research Report No. 17.

Cowan H.J. 1966, An Historical Outline of Architectural Science, Elsevier, Amsterdam.

Crawley, D.B. et al. 1996, ‘Workshops on Next Generation Building Energy Simulation Tools’, ibpsaNEWS, Vol8 No. 1, pp.2-6

Crawley, D.B. et al., ‘Workshops on Next Generation Building Energy Simulation Tools. Part 2: Contrasting Developers and Users’, ibpsaNEWS, 1997 Vol9 No. 1, pp.1-5

Howard, H. 1960, ‘Pipe Sizing with an Electronic Computer’, JIHVE, Vol 28, No. 1, pp23.

IES, 1997, SunCast: Solar Mapping and Insolation Studies, Computer Software Manual, Integrated Environmental Solutions Limited (IES), Glasgow.

Lomas, K., Eppel, H., Martin, C. And Bloomfield, D. 1994, ‘Empirical validation of thermal building simulation programs using room test data: Volume 1’, Final report of IEA Annex 21: Calculation of Energy and Environmental Performance of Buildings.

Longmore, J. 1968, BRS Daylight Protractors, Department of Environment Building Research Station, Her Majesty’s Stationary Office (HMSO), London.

Lynes, J.A. 1968, Principles of Natural Lighting, Elsevier Publishing Company Ltd, Great Yarmouth, England.

Manning P. 1987, ‘Environmental Evaluation’, Building and Environment, Vol22 No. 3, pp.201-208

Palmer, G. 1997, ‘Cool Running’, Building Services Journal, October ‘97.

Papamichael, K. 1997, ‘Designers and Information Overload: A New Approach’, Advanced Buildings Newsletter, Vol 1, No. 18, Royal Architectural Institute of Canada.

Parsloe, C.J. 1995, ‘Over-engineering in Building Services: An International Comparison of Design and Installation Methods’, BSRIA Technical Report No. TR21/95, BSRIA.

Phillips, R.O. 1983, Sunshine and Shade in Australasia, Australian Government Publishing Service, Canberra.

Race, G.L. 1997, ‘Design Margins in Building Services’, Building Services Journal, August ‘97.

Owen, J. and Roy, G., 1989, SR: Shadow and Reflection Analysis, Computer Software Manual, School of Architecture, The University of Western Australia, Perth.

Ruyssevelt, P. and Batholemew, D. 1997, ‘Stimulating Simulation, Computer Tools in Building Services’, Building Services Journal, October ‘97.

Schroeder, M.R. 1965, ‘A New Method of Measuring Reverberation Time.’ J. Acoust. Soc. Am. Vol. 37, pp 409-412.

Spencer, J.W. 1965, Calculation of Solar Position for Building Purposes, CSIRO Division of Building Research Technical Paper No. 14.

Szokolay, S.V. 1980, Environmental Science Handbook, The Construction Press Ltd, Lancaster, England.

Szokolay, S.V. 1987, Thermal Design of Buildings, RAIA Education Division, Canberra.

Szokolay, S.V. 1987, Archipak Version 1.1, Computer Software Manual, University of Queensland Dept. of Architecture.

Szokolay, S.V. 1996, Solar Geometry, PLEA Note No. 1, University of Queensland Dept. of Architecture.

Vorlander, M. 1989, ‘Simulation of the Transient and Steady-State Sound Propagation in Rooms Using a New Combined Ray-Tracing / Image-Source Algorithm.’ J. Acoust. Soc. Am. Vol. 86, pp 172-178.

Ward, G. 1994, Supplementary Notes - Radiance Synthetic Lighting Simulation Application, Lawrence Berlkey Laboratories.