Material Computation Method For Nonstandard Materials Based On Machine Learning

In non-linear architecture,form finding and fabrication are inseparable.But in the process of actual architectural design and fabrication,the design algorithm usually does not involve material characteristics.So the process often leads to errors and structural redundancy during construction.Materials,with the standardized performance,exist in architecture.aims at this problem,Material programming propose the approach which is materials lea ding the process of design and construction and maximizing the material performance.But the existing material computation method still has a great deal of manual intervention in the process o f research.Artificially definition and lots of similar experiments of the material can not achieve very good expectations.At the same time,As most materials which studied in material computation is nonstandard,there are a lot of errors in the process of artificially defined in.This paper present a new materials computation system based on machine learning which is made by artificial intelligence.Through the given sample data,computer model can independent learning behaviors of materials and give suggestions during architectural design and construction.In this paper,material computation based on the machine learning method is adopted.Multilayer nonlinear abstraction of material properties is carried out by artificial neural networks.And based on the system,article present a design practice and model production Through the feasibility analysis,we explore the application of the system and the existing shortcomings of the whole system.And put forward a reasonable outlook including the combination of machine learning and material computation,and the combination of machine learning and architectural design,.