A new academic complex for Leeds Metropolitan University gave Feilden Clegg Bradley Architects the opportunity to use computational modelling to optimise the building's facade design and even out daylight levels.
Feilden Clegg Bradley Architects' new academic complex for Leeds Metropolitan University, which will provide 10,000m 2 of offices and teaching spaces for four departments, including architecture, was conceived as a solid landscape form, drawing on Yorkshire's rich geological heritage. The scheme, which was granted planning permission in March 2007, also includes a new Baptist church, a café and exhibition space, and 240 student bedrooms/studios.
These will be organised in two buildings of irregular massing and varying heights, ranging from three storeys adjacent to low-rise listed buildings to a maximum of 23 storeys. Rainscreen cladding of Cor-ten steel was selected as a solid, sculptural and weathering material, punctured by cascading glazing inspired by water owing through a rock formation.
Early facade studies - using randomised, full-height glazing panels of varying width - explored the use of more glazing at the lower levels and progressively less transparency higher up.
This was because upper storeys typically have greater access to light and so need less glazing, whereas lower levels, with more overshadowing from nearby buildings, need more glazing to achieve an equally bright interior. For ease of construction and to reduce costs, the cladding was rationalised to a grid of 1.5m-wide panels, and the facade design was developed through a process of iterative computer optimisation by modelling overshadowing and solar orientation on each building elevation. This computational modelling allowed us to test and quantify our design intuition and develop the design accordingly.
Light levels were compared on the basis of 'average daylight factor'. Recommendations vary, but less than 2 per cent is likely to require artificial lighting and 5 per cent is generally perceived as very well lit. For the Leeds academic complex, a target of 3 per cent was agreed in order to provide adequate natural light, avoid overheating and maintain the desired solidity of the exterior.
An initial analysis of the amount of glazing needed on different oors and in different facades to achieve the 3 per cent average daylight factor showed that there was more variation horizontally around the building than there was vertically. The vertical gradation in the percentage of glazing needed was only evident on areas of the facade in close proximity to other buildings.
Since the team was keen to progress the facade design on this basis, detailed data on overshadowing was commissioned from the BRE.
This gave us a 'theta' value (a measure of the area of sky a window can receive light from) for every 1.5m module on every oor of the building, from which it was possible to calculate the optimum percentage of glazing for that module. This data was colour coded and applied as a scaled facade to a 1:500 model.
Overheating and orientation were studied in a similar modelling exercise. For the Leeds area, Part L recommends limiting heat gains to 41W/m 2 of oor area, counting only oor area within 6m of the facade. An assumption of 21W/m 2 for internal gains left a maximum of 20W/m 2 for solar gain. This detailed data was again converted into a coloured spreadsheet 'facade'.
Comparison of the two sets of data showed that substantial areas of the building would be subject to overheating to achieve the desired daylight factor. The use of solar glazing - which transmits only about 40 per cent of solar gain compared to 70 per cent for typical double glazing - meant that it was possible to keep the required glazing around most of the building.
All calculations and analyses for the modelling were performed using Excel spreadsheets. With in-house computing expertise, this hard data was converted into a facade design using a Visual Basic for Applications (VBA) program within Excel.
VBA was preferable to a standalone application, which would have required more extensive programming and might never have been used on another project. The VBA program could also be imported easily into Microstation, our primary CAD package.
The algorithm itself is a set of recursive conditional statements, making weighted decisions based on the numerical daylighting analysis and its design, at its current iteration. The user can define the way the algorithm responds to changing design requirements, using variables. The facade was divided into groups of four 1.5m modules. Each group was averaged to determine the amount of glazing required and assigned a number of glazed panels accordingly, which were randomly placed by the program.
The randomised placing was then refined for aesthetic reasons, because the intention was that the Cor-ten - denser at the top of the building - would appear as if it were being weathered, thinning out as it came closer to the ground. The algorithm arranged the solid and glazed modules based on the surrounding panels, so that the probability of placing solid panels below or diagonally below other solid panels was increased. This created links between the solid panels at the top of the facade and those further down, giving the appearance of a 'cascade' of glazing down the facade. This was executed with a single button in Excel, meaning that many options could be created rapidly. Although it would have been possible to link the algorithm within Excel directly to Microstation to initiate the drawing, this was quick to achieve manually and time pressure meant that we opted for the safer route.
As a progression of these ideas, we have used radiosity renderer Maxwell to simulate the amount of daylight illuminating the proposed facades on another project. Renderings were made throughout a 24-hour period during summer and winter solstices, effectively giving a year-round analysis. A standalone program was then built to analyse the brightness of these renderings. The results replicated the daylighting analysis provided by the BRE for the Leeds academic complex, indicating a way forward for this type of analysis in-house, and bypassing many of the sums.
Andy Macintosh, Richard Priest and Alex Whitbread are, respectively, architectural assistant, architectural software engineer and partner at Feilden Clegg Bradley Architects.