Our project aims to determine suitable workstation arrangement for office typology in a conservation shophouse in Singapore through maximise daylight on the worksurface to achieve 300 lux (to a maximum of 3000 lux) for good lighting condition. In a conservation shophouse, where the building envelope and facade cannot be altered. There are two daylight sources into the space through the front facade windows and either back facade windows or light shaft, getting good daylight depends on the location of workstations.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide2.png)
Interior designer can use the tool to easily predict lux level on workstation to layout desk configurations to maximise daylight in the early design stage. The project could expand to more shophouse typologies and make it into an early stage design app for interior designers. Alternatively, it could also be used for modern office layout as well.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide3.png)
Our project methodology consists of developing the dataset through running parametric models and daylight analysis to compile a data set for machine learning. We run simple annual daylight simulation and plot the analysis data, then we come up with desk configurations for pulling out daylight levels at each desk. At the end, we have over 20,000 data sets to train our model. We export our data as CSV file for the machine learning process.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide4.png)
Below is our GH process for daylight calculation.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide5.png)
This is our table configuration to pull out the data for machine learning.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide6.png)
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide7.png)
We train the model through machine learning process. We are able to correlate back the training data to ensure that the prediction model came out properly.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide10.png)
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide13.png)
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide14.png)
When we attempt to plot back the daylight level to our 3D model environment. We have to separate the model into front and back zone in order to get the proper input data to display.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide11.png)
Our user interface is through Rhino 3D model environment, Grasshopper, and python components. At the end, we are successfully to pull out the training data based on where we place desks in the model.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide15.png)
Further outlook, we have achieved what we set out to do although in a simple way, we could apply the same method for other shophouse typology and develop a front app to run a simple daylight study for early stage design. Additionally, we could later develop a portal for designers to input their own floor plates to run daylight and predict lux for each workstation configuration. This can be a very useful tool for early stage design process.
![](https://blog.iaac.net/wp-content/uploads/2024/07/Slide16.png)