Machine Learning and Optimization Techniques for Steel Connections
1. Machine Learning and Optimization techniques for
Steel Connections
Lorenzo Greco, www.parametricism.co.uk, AKT II
Paper number
205
lorenzogreco@gmail.com
2. Sberbank - Zaha Hadid, 200000mq free form
Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com
3. Random boring factory - unknown
Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com
4. Most amazing city - many
Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com
5. Sample model - Revit
Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com
12. Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com 12
Optimized
dataset
Accuracy benchmark
Robot's
analysis
Real
case
Real case
with
optimized
joints
13. Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com 13
Breakdown of utilization factor. Blue: White City Place Red: ML + Optimization
Comparison: ML vs ground truth
14. Lorenzo Greco, parametricism.co.uk | AKT II, lorenzogreco@gmail.com 14
● Automize joint design and steel design embedding engineers’ decisions
● Gain insights from patterns emerging from real projects and incorporate them in code and guidelines
● Automate checks on projects by proof engineers and city council
● Retrofitting
● Simplify explorations for complex structures
● Expand to MEP layout, architectural finishing, detailing, etc...
Conclusion and future progress
Notes de l'éditeur
Data distribution and correlation matrix
White City Place, a 7 storeys building in London designed by AKT II with Allies and Morrison architects with Lend Lease and commissioned by Stanhope. The building is situated north of White City underground station and is part of the redevelopment of the BBC Media Village.
We want to maximize the area under the graph. This shows the accuracy against recall(UF ratio)
(a) Accuracy as function of absolute error in predicting the correct utilization factor (UF), for a K-Neighbour Regressor trained on an optimized dataset
(b) Accuracy as function of absolute error in predicting the correct utilization factor (UF), for a K-Neighbour Regressor trained on an optimized data set, derived from Robot's analysis
(c) Accuracy as function of absolute error in predicting the correct utilization factor (UF). The algorithm is tested against a real case.
(d) Accuracy as function of absolute error in predicting the correct utilization factor (UF). The algorithm is tested against a real case, but with the goal of hitting the optimized algorithm that would fit rather than the actual ones used.