Multi Constraint Layout Of High-rise Buildings And Its Wind Ambient Intelligence Intelligent Prediction Method

With the acceleration of urbanization and rapid economic development,the low efficiency of land resource utilization has become a bottleneck that restricts the sustainable development of cities.The design work of traditional architectural layout planning requires a lot of manpower and time,mainly relying on manual provision of various architectural layout plans.Therefore,studying the method of automatic generation of building layout can effectively alleviate the design pressure of architectural designers and provide decision-making support for regional development planning and future urban design.This automatic generation method can achieve intelligent design,reduce human errors in the design process,and improve design efficiency,thereby better meeting the rapid development needs of urban planning and design.The current research on automatic layout of high-rise buildings has explored various algorithms.Specifically,the time required to generate a layout that meets the specifications is enormous,and even if it meets the specifications,the generated layout may be abandoned due to not fully showcasing the beauty or elegance pursued by the architect.In addition,while computational fluid dynamics(CFD)can be used to simulate the building wind environment when conducting wind environment assessment after building layout is formed,its use process is quite time-consuming;At present,the method of using data-driven generative model to predict building wind environment can solve the efficiency problem of CFD simulation of wind environment.However,due to the lack of extraction of specific image features of building wind environment,the generation quality cannot fully meet the requirements.To address these issues,this article proposes a data-driven case recommendation algorithm for high-rise building layout generation and a wind environment prediction algorithm based on spatial domain enhanced pix2 pix.The specific research content includes:Ⅰ.A case recommendation algorithm for high-rise building layout generation based on reinforcement learning and pattern clustering is proposed to address the problem of automatic layout of high-rise buildings under multiple building constraints.This algorithm combines the normative requirements of building design,including building fire prevention,sunlight,building property lines,etc.,and designs corresponding reward functions.Interact building units as intelligent agents with the environment,adjust the position of building units according to regulatory requirements,and continuously learn and optimize to ensure that the generated layout scheme meets the requirements of building regulations.At the same time,this algorithm can output multiple layout schemes that meet the set constraints,providing architectural designers with more choices.In order to further improve design efficiency and quality,the algorithm also performs pattern representation and automatic clustering based on Fourier descriptors on the building clusters in the generated scheme,providing redundant and diverse typical design patterns.Finally,the feasibility and effectiveness of the method were verified in a real environment of a certain plot in Beijing.Ⅱ.Aiming at the wind environment prediction of high-rise building layout,this paper proposes a method of wind environment prediction based on spatial domain enhanced pix2 pix model.Firstly,the building form and pedestrian wind environment are represented as images,thus transforming the problem of building wind environment prediction into an image conversion problem.Secondly,the low frequency and high frequency information of the image are extracted by Gaussian fuzzy,and the L1 distance is calculated respectively,and the sum is added to the generation model as a loss term.By limiting the high and low frequency information separately,the key areas of wind speed and airflow can be better captured.The proposed method is applied to the wind environment simulation data set of residential areas in Beijing.Experiments show that the accuracy of the existing adversarial network model is improved through spatial domain enhancement.In the evaluation of model generalization ability,adding space loss term can improve the performance of various original generated models.In addition,the building wind environment prediction time has been reduced from hours to milliseconds.Finally,a framework for automatic generation and analysis of high-rise residential building layouts has been constructed,which mainly includes three modules: building layout generation,building layout clustering,and building layout wind environment prediction.Finally,predictions and analysis were conducted on the solutions that meet the architectural constraints and eliminate redundancy.