Exploration On The Morpholygy Analysis And Intelligent Generation Method Of Building Complex

With artificial intelligence technology sweeping all walks of life,it empowers computers with the ability of autonomous learning.The existing research on architectural generative design based on in-depth learning technology focuses on the space types with relatively simple layout,such as residential areas,youth apartments and schools.They simplify the programming process of traditional generative design and demonstrate remarkable results.However,due to the complexity of morphological form and the limited number of samples,the design task of other public buildings such as commercial service buildings is still a blank field to give full play to the advantages of artificial intelligence technology.Therefore,using computers to realize the form layout automation of public buildings in order to improve work efficiency and help the upgrading of the industry has become a topic worthy of research.Taking the commercial service building complex as an example,this study hopes to distinguish the morphological types of the building complex scientifically and efficiently by quantifying and analyzing the form of the building complex and using the mathematical laws reflected by the form index system.On the basis of this morphological database,through deep learning and other computer technology methods,the layout form of the commercial service building complex at the plot scale is generated and designed to help designers quickly obtain the results of the scheme for reference.The research content of this paper includes four parts: background and theoretical basis,construction of technical method framework,quantitative analysis and classification of morphological features,experiment and review of morphological intelligence generation.The first part combs the relevant research on the automatic generation of architectural space form under the wave of artificial intelligence from the two aspects of morphological typology theory and morphological intelligence generation method,analyzes the existing research results,emphasizes the lack of current morphological typological analysis ideas,the weakness of design thinking and poor interpretability of deep learning technology.The article clarifies the necessity of intelligent generation to open up the connection from morphological type analysis to intelligent generation under the fundamental principle of learning and reference,and builds a design thinking framework from morphological type division under index constraints to intelligent generation design aided by deep learning.This serves as the theoretical foreshadowing of the later generation practice.The second part,on the one hand,introduces the method of classifying the morphological types of building clusters based on the control function of indicators,including quantitative indicators and calculation formulas of building clusters,extraction of the core index system for controlling the morphology,two-step clustering algorithm,specific method of initial classification of shape types and subset subdivision.On the other hand,compare the existing mainstream deep learning models,summarize their characteristics,and explore the adaptability of each algorithm to the research topic of this paper,so as to choose Pix2 pix as the research model.Then it introduces the technical points of data format standard conversion,experimental platform,model network structure and generation scheme vectorization in detail.This part builds a technical route for the whole-process intelligent generation.The third part,under the guidance of the second part of the technical framework,takes Nanjing as a case to choose commercial service plots and building data sets.The data samples include attribute labels such as land useage,plot shape,plot area,building area and building floors,and calculate other morphological indicators.Based on this,the article uses excel,spss and other tools to carry out objective statistical analysis on the morphological characteristics of commercial service building blocks and building entities.From the mathematical structure of spatial morphological indicators,the morphological characteristics of commercial service buildings are quantitatively analyzed,and at the same time,principal component analysis and clustering algorithms are combined to divide the refined morphological types to provide the generation direction and data sample basis for the deep learning algorithm model.The fourth part is to build and train the algorithm model.Based on the type sample database cleaned and processed in the third part,the training experiment of unconstrained morphological samples and the training experiment of distinguishing morphological samples are carried out,in order to obtain a guided deep learning model.Further,take Xiyan Lake in Wuxi as an example to test the effect of the model in the specific project practice,analyze the rationality of generating the morphological layout scheme of commercial service buildings from the perspective of planning discipline,evaluate the effectiveness of the generation model,explore the factors affecting the effectiveness of the algorithm model,and summarize the potential optimization direction of the model.The final research results show that,based on the theoretical knowledge of urban planning disciplines,using clustering algorithms to divide refined form types can further promote deep learning technology to efficiently and automatically generate building complex forms that meet the needs of designers.Although there is still a gap between the current generation results and manual design,it is enough to foresee the infinite potential and possibility of artificial intelligence in the planning and design of urban spatial form.