Developing structural design automation programmes

Jana Miller, Bentley Systems 

Hyundai Engineering leverages 3D modelling and AI to automate design of civil and architectural structures

TO minimise time spent on simple repetitive engineering tasks and allow engineers to focus on quality and creative engineering, Hyundai Engineering initiated a project to develop an automated design system for plant steel structures.

In particular, the aim was effectively respond to frequent design changes by developing a programme for design information exchange between relevant departments and automatic designs for steel frame construction planning of chemical and electric power plants.

The focus was on shelter and pipe rack designs for these plants, seeking to streamline information exchange and automate design processes using AI and machine learning.

As the plant design process typically involves multiple engineering changes throughout the design stage, revisions to the shelter and pipe rack designs are necessary to meet the modified plant design specifications.The client designed shelter structures that are one to three levels high where various mechanical devices are temporarily or permanently located in plant projects.

Pipe racks are raised structures used in industrial facilities to support pipes, conduits and cable trays.

Both elements play a critical role in industrial plant design.

As the plant design process typically involves multiple engineering changes throughout the design stage, revisions to the shelter and pipe rack designs are necessary to meet the modified plant design specifications.

To help simplify and accelerate structural design methods, as well as meet evolving demands, and other emerging industries, there is ongoing AI-based design automation technology research and development.

Manual workflows, collaboration, and change management

The general structural design workflow for industrial plants typically involves engineers using a structural analysis programme and manually entering design conditions and loads to conduct structural analysis. This process involves time-consuming, manual data entry for the application and calculation of structural loads that is error-prone, resulting in countless design changes.

The process is very tedious and complicated, requiring enough time to review human errors, which leads to an increase in task burdens that are amplified along with the change of designs.

Standardisation also becomes difficult due to the need for subjective judgment of designers, depending on the structure and the varying design criteria for each project. The lack of standardisation and frequent design changes result in increased time and costs to deliver a project.

The general structural design workflow for industrial plants typically involves engineers using a structural analysis programme and manually entering design conditions and loads to conduct structural analysis. This process involves time-consuming, manual data entry for the application and calculation of structural loads that is errorprone, resulting in countless design changes.The client realised that to optimise change management and industrial plant project execution, it required crucial collaboration among piping, electrical, instrumentation and mechanical disciplines.

Collaboration is important because the information derived from these disciplines substantially contributes to the structural design aspects for civil and architectural infrastructure.

Constant revisions lead to registrations of design information (e.g., pipe weight) and examinations and updates of the changes from multiple teams; sometimes, the same work [had to] be repeated.

The smart technology team sought to standardise and streamline plant engineering workflows by using intelligent digital solutions to cohesively integrate diverse disciplines, automate designs and eliminate repetitive tasks for steel frame construction planning of industrial plants.

Advancing STAAD application with machine learning

STAAD was selected to integrate 3D modelling and AI, automating and streamlining design workflows and creating designs through machine learning predictions. This began with developing an automatic design system for steel structures, focusing on pipe racks and shelters. The design process encompassed structural components, such as cables and trusses, as well as structural member connections.

Therefore, the team was required to integrate data, such as weight, for the plumbing, mechanical and instrument teams into the 3D model. An AI-based algorithm was incorporated into STAAD that eliminated manual data entry and automatically imported the weight and/or other relevant data for the different disciplines into the model, creating an automated design programme.

When such data is synchronised to the automation programme, the weights, such as pipe load and cable tray load, are applied automatically to the STAAD model. The design work that was conducted by the individual engineer could then be digitalised and managed through a unified best practice.

The AI-based structural design begins with automated shelter modelling, and progresses to database construction which can be tailored according to structural dimensions and culminates by predicting recommendations for trusses, rolled beams, and/or column distances.

Prior to automation, the shelter’s roof structure was determined based on existing manual design estimations. However, after the automation process, countless scenarios can be digitally examined to determine the optimal roof system.

The 3D modelling and AI automation design programme that the client developed incorporates automation, prediction, and optimisation processes into plant engineering practices, expanding the application through machine learning and actively utilising digital technology from two engineering perspectives.

First, the design variables and results from the automation are established as a database and run for machine learning to be expanded as an AI design. Second, the design information that was exchanged as drawings and documents between relevant departments is now under unified management in the form of 3D model data.

Smart technology drives savings and standardisation

Now, engineers can determine their structure systems based on the database instilled through practice instead of relying on personal experience and senses.The client’s newly developed smart design system for civil and architectural plant structures increased work efficiency, reducing the time spent by engineers on manual, repetitive tasks.

The automated programme saved design variables in a digital database, which, when combined with machine learning technology, led to the creation of a true AI design with the ability to produce identical results consistently through repeated tasks.

It means that when an ultimate AI design is realised by using machine learning, the results can be deduced faster than when the engineer performs the task manually.

The result itself was a precise, integrated solution of 3D modelling with intelligent digital workflows that provide accurate design information and accelerate front-end engineering (FEED) design by at least 30%. It also optimised the volume of construction, reducing the construction design costs by more than 20% by eliminating design-time errors.

Based on 50 bidding projects, 25 FEED projects and 20 execution projects over five years, Hyundai Engineering estimates saving almost USD$2m in outsourcing costs.

The benefits of the client’s smart technology solution not only lie in accelerated workflows and cost reduction, but also in its capability for expansion and standardisation, creating AI designs through machine learning predictions that can be used in future projects.

Using the shelter AI-design automation programme, the team created a database of 1,680 scenarios, generating 27 million prediction models. For pipe rack design, weight and geometry data within drawings and documents can be unified through a 3D model in a digital database, leading to streamlined, standardised and intelligent design processes.

Now, engineers can determine their structure systems based on the database instilled through practice instead of relying on personal experience and senses. 

Jana Miller, Bentley Systems

www.bentley.com

@BentleySystems

All images courtesy of Hyundai Engineering.