Introduction

Automating aspects of the design process of large civil infrastructure work could improve design quality and maximise user satisfaction levels. In this project a new algorithm based on machine learning and neural network will be developed for deep searching the optimal design configuration for large infrastructure works. The research outputs will enable better design for urban infrastructures (e.g. communications, energy, water and transportation) in terms of safety, economic viability and sustainability.

Explaining the science

The current design process for urban transportation network and large infrastructure projects relies on manual data collection from various aspects, such as building codes, laws and owner requirements. An initial project design is generated based on these data without an effective mechanism for verifying whether the design configuration has fulfilled all the constraints, and whether it remains optimal in both safety and economic expectations.

The new proposed algorithm will mimic nature’s evolutionary approach to design. It starts with the setting up of design goals and then explores innumerable possible permutations of solutions, while simultaneously considering all constraints and requirements for finding the best design configuration. The algorithm will execute the design cycle multiple times and only applicable and near-optimal solutions are preserved for each iteration.

The data regarding the constraints and requirements of urban network and large civil infrastructure projects are large in volume, and cannot be used directly. This work aims to establish a systematic and effective process for data extraction from the building information models (BIMs). In addition, this research plans to develop an agent-based infrastructure simulation model for transportation infrastructure networks, which identifies the interactions and interdependencies between urban infrastructure systems and autonomous vehicles.
 

Project aims

Urban civil infrastructure design is the process of planning future systems and services for 
the greater good. As the size of the city grows, the design of the infrastructure projects become increasingly complex and interconnected. In this context, the complexity of engineering systems is growing faster than the design engineer’s ability to address it. 

Therefore, multidisciplinary engineering teams need to work together, integrating diverse expertise across disciplinary models. This collaborative work can potentially benefit from rich sources of data and an automated design process which can analyse all the  feasible design configurations based on the abundant input data. Thus, it's necessary to develop a computing algorithm which is able to identify critical features of design problems and automate routine design processes.

The aim of this project is to establish this kind of algorithm for engineering design, improving the overall process of design decision-making in highly uncertain and complex engineering and commercial contexts. The developed algorithm will be capable of leveraging extensive datasets, and integrating real world data across the entire project life-cycle.

The output of this research will assist the UK industrial strategy which sets out a commitment to accelerating change in the infrastructure and construction sector through a 'transformative programme'. It will do this by bringing together construction, digital technology, manufacturing, materials and energy sectors to develop and commercialise digital and off-site manufacturing technologies.

This project is part of the Data-centric engineering programme's Grand Challenge of 'Data-driven engineering design under uncertainty'.

Applications

This research can be applied to design infrastructure systems (e.g. transportation, water, electricity or or communication) which possess better resiliences toward undesired events. The study can also assist the construction of new mobility platforms that are emerging, such as autonomous vehicles, and how they may impact urban engineering initiatives. 

Resilient infrastructure systems are able to maintain their function even when adverse events occurred (e.g. natural disasters or congestion). Thus, enhancing the resilience of infrastructures by improving their capacity to resist, recover and adapt are major concerns which are necessary to consider during the design process. Through the proposed data-driven algorithm, the engineering design in infrastructure systems can be optimised for multiple criteria, achieving high resilience and economy at the same time.

The concept of shared autonomous vehicles could significantly reduce the number of cars in future cities and therefore improve traffic safety, mitigate congestion, increase energy efficiency and fundamentally change the nature of urban mobility. The planning of a new urban mobility system requires seamless coordination between vehicle capacity, vehicle repositioning strategies and the location of the recharging stations. This research will build upon recent meso-level (between micro- and macro-) urban modelling work, focusing on the interaction between different transportation modes. It will result in the development of a virtual urban testbed that will be used to trial new mobility concepts, facilitated by coordination and urban systems integration. An agent's behaviour within the testbed, and system interaction patterns in the model, will be validated using real world data.

Recent updates

The current design process for urban transportation network and large infrastructure projects relies on manual data collection from various aspects, such as building codes, laws and owner requirements. An initial project design is generated based on these data without an effective mechanism for verifying whether the design configuration has fulfilled all the constraints, and whether it remains optimal in both safety and economic expectations.
  
The new proposed algorithm will mimic nature’s evolutionary approach to design. It starts with the setting up of design goals and then explores innumerable possible permutations of solutions, while simultaneously considering all constraints and requirements for finding the best design configuration. The algorithm will execute the design cycle multiple times and only applicable and near-optimal solutions are preserved for each iteration.

The data regarding the constraints and requirements of urban network and large civil infrastructure projects are large in volume, and cannot be used directly. This work aims to establish a systematic and effective process for data extraction from the building information models (BIMs). In addition, this research plans to develop an agent-based infrastructure simulation model for transportation infrastructure networks, which identifies the interactions and interdependencies between urban infrastructure systems and autonomous vehicles.

Organisers

Researchers and collaborators

Funders