Digital technology is constantly present in our daily life and is a main contributor to the changes and developments in our society. Technology opens a wide range of possibilities, many of which were not achievable in the past. Technology devised for one particular purpose is usually captured and adapted for other purposes and industries.

The tunnelling industry seeks to adapt and implement technology in order to improve the accuracy, productivity and efficiency of its activities, including design, procurement, supply chain, construction, commissioning, operation and decommissioning. Many examples of this can be seen, such as: more effective data management and digital delivery; use of robotics to mechanise manual activities to improve health and safety; and instrumentation and monitoring or advanced mathematical and statistical models to manage the financial risks of a tunnelling project.

The purpose of this paper is to describe and explain the advantages and challenges of using parametric design in tunnelling. This is an existing and widely used method in many fields, such as architecture and mechanical engineering, but adapted to the conceptual and detailed design of tunnels and shafts.

The paper explains how I developed a system to apply parametric design to a live project. I will explain what it is, how it was developed, the benefits it brought, and the lessons learnt for future applications.

This methodology was applied to the design of the High Speed 2 (HS2) Euston Station Enabling works – Substation and Ventilation Shaft project, one of the first parts of HS2’s infrastructure to be built in London, and which will allow the new station works to commence. The project for Costain Skanska Joint Venture (CSJV) with Mott MacDonald as a main designer, will build a new ventilation and substation to be connected to the existing infrastructure at Euston Station.

The most significant innovation I brought to the project was to develop and apply a fully automatised geometry modelling process, which was used for the first time at Mott MacDonald for a fee-earning project.

PARAMETRIC MODELLING

Parametric modelling is a method that allows the building of a 3D model in a more efficient way. The conventional method consists of manually creating and editing, throughout the project development, the project geometry. With parametric modelling, the geometry is coded with algorithms (rules) and then, the 3D model is controlled by a series of inputs (parameters) which can be easily changed. When the inputs are modified, the geometry algorithm is re-run and the model is automatically updated.

Parametric modelling is, therefore, a process built upon a series of rules that generate a geometrical model based upon a set of pre-defined parameters which govern the design. This enables some historically time-consuming tasks to be automated and applies changes in a more dynamic manner.

In the example presented in Figure 1 below, a three-dimensional surface is coded, and the shape is controlled by the parameters of two curves in the space. When any of those parameters are changed, the surface changes accordingly.

Applying changes to a 3D model after every design decision can retard the progress of a project. This may impact the programme, consume additional resources or cap the potential changes and improvements due to the lack of time.

Checking the model of a complicated 3D element is also challenging and sometimes defining a process or routine for this is very difficult. The checking process is difficult to track, relies on the subjective view of the checker, and often taking measurements or angles on curved 3D surfaces can become a quite unmanageable task.

Therefore, the main objectives of parametric modelling are to increase the efficiency of the modelling process; allow the optimisation process to be more agile; generate a data assurance workflow; and increase reliability in the model by minimising errors.

As part of its digital delivery and technical excellence initiatives, engineer Mott MacDonald carried out a series of research activities in order to explore the applicability of parametric modelling to tunnelling design. I was tasked with exploring the various alternatives and to investigate the different software available to achieve the purposes described.

Autodesk Dynamo and Bentley Generator were found to be equivalent software for parametric design. They allow the coding of rules, and even the ability to generate sub-routines or customised nodes where specific tunnelling aspects could be defined. Autodesk Inventor, even though not strictly a parametric modelling software, has been widely used in mechanical engineering and more recently by tunnelling manufacturers; it has also been found to be suitable to parametrise elements, such as precast concrete segments that have a high level of detail. This application is really valuable for the production of a detailed 3D model of segments that form part of a TBM tunnel ring, which is increasingly becoming a main deliverable of a TBM project.

APPLICATION CASE

The Euston Station Ventilation and Substation Enabling Works is a multidisciplinary project that involves the design and construction of a new headhouse, shaft and tunnels that will connect into existing infrastructure at Euston Station. The existing headhouse is located within the footprint of the new HS2 Euston terminus station. It includes the excavation of a 13m-deep basement and a substantial oversite development. This project will need to be commissioned so the HS2 station excavation works can be completed.

In order to connect the existing cable and ventilation galleries from the headhouse to Euston Station, a set of mined tunnels are to be constructed below the HS2 Euston station.

The tunnelling part of this project can be seen in Figure 2 (left) and consisted of:

  • A temporary shaft consisting of a segmental lining in the upper part of the shaft and an enlarged, sprayed primary lining and cast in-situ secondary lining on the shaft’s lower part.
  • Two connection tunnels to be constructed using sprayed concrete primary linings and cast in-situ secondary linings, linking the temporary shaft with the new headhouse and with the existing infrastructure
  • Two connection adits which link the new tunnel to the existing assets.

My team and I reviewed the inherited RIBA3 design (Figure 3) and advanced the design into RIBA4. We suggested several changes to facilitate the construction activities and construction programme. The two connection galleries to the existing assets arrangement were changed into a single short tunnel and two short adits. The temporary shaft lower part was enlarged to accommodate the larger openings (Figure 4).

The temporary shaft will be the first element to be constructed with both connection tunnels being excavated from the shaft.

Only 1-2m of cover distance separates the HS2 Euston station basement base slab level to the crown of the new tunnels. The reason for such a low cover is due to the current location of the existing assets of the LU station from where the cables and ventilation ducts will be ultimately connected to the new ventilation and substation headhouse.

One of the main challenges of this project was dealing with interfaces, such as the more than 100-yearold array of galleries in the existing infrastructure of Euston Station; and the excavation of the HS2 Station, where the loads transferred from base slab to tunnels changed during the detailed tunnel design.

Due to the intricate geometrical alternatives considered and the time taken to get the modifications introduced into the BIM 3D model and the design process, I considered that a solution had to be sought. This would aim to automatise the process and rapidly make changes to the model so that design of options could be carried out timely and without impacting the design or the construction programme.

The project size and the problem encountered led us to decide to use the explored methodology of parametric design in this project.

DEVELOPING THE PARAMETRIC MODELLING PROCESS

In this project, I was responsible for creating and developing the strategy and the tools of the fully automised modelling concept and its subsequent implementation. I divided the design process into eight steps, which were as follows:

1. Design and sketches, as usual

The conceptual design is one of the main and most important activities. The selection of the design solution remains the same as in the traditional design workflow. This is a stage where space-proofing requirements and the structural and geotechnical calculation approaches are also defined.

2. Identify and log the driver parameters

When the design concept is defined, the parameters that drive it are identified. For a tunnel design project, the typical parameters would be coordinates, lengths, thicknesses or radii.

The parameters are logged into the input database which will be the ‘single source of truth’. This database registers the design drivers, the date they were updated and the status of that parameter in the design (‘Work in progress’, ‘fixed’ or ‘to be updated’). The input database is stored in a common digital environment so the team can access it with certain permissions. The input data contains not only geometrical parameters but also other design parameters such as coordinates, thicknesses and our tunnel radius. This will be a valid source of information for the geometrical model, the calculations and the design report.

3. Coding the geometry and generating tunnelling tools

The geometry is coded in Autodesk Dynamo in a script file and ranges from generating simple points and lines, to more complex geometry. As a result, a fixed, but editable if needed, algorithm for each asset is to be modelled in 3D.

Sub-routines, which will be used repeatedly, are created and used throughout the script. As an example, I created the Tunnel SCL profile ‘custom node’, where six parameters define the outline of the tunnel. Those sub-routines can be written in Python or directly in the Autodesk Dynamo coding language.

The coding process is considered as ‘user friendly’ and very intuitive since the software used in this case, operates with similar commands to AutoDesk AutoCAD, a widely-known CAD software.

4. Linking the coded script to the input database

The algorithm is built in such a way that the drivers of the design (inputs) are called from the Input database, which can be a spreadsheet or other type of file. That effectively links the ‘source of truth’ with the algorithm that defines the geometry.

5. Generate the 3D model geometry

The algorithm is run in the software. It calls the inputs and provides a geometry file as an output. Other data outputs can be obtained, such as volumes of different parts of the asset or lists of logical information previously coded.

6. Model validation process

I generated a model validation and model-checking workflow to ensure that the code worked correctly. The goal is to ensure that the inputs provided match the 3D model created.

The script is validated in each version. Additional 3D checks are made to the geometry, but this process is much simpler when we have confidence that the script provides the correct geometry, provided the inputs. The primary check changes to become a list of parameters instead of a complicated 3D model.

The workflow, together with the version locking, provides a trackable and reliable assurance process.

7. Model and drawings generation

Once the 3D model is created in Autodesk Dynamo, then checked and the script validated, the geometry is imported into the HS2 common environment software MicroStation. At this point, the engineer has effectively delivered their design to the BIM technician, who may only need to make small adjustments to the model if necessary. Out of that 3D model, the appropriated 2D cuts are extracted in order to produce the drawings.

8. Redesign

If a design decision is taken, and the geometry needs to be modified, by updating the input database and then re-running the Autodesk Dynamo Script, the model will be automatically updated, providing a more agile management of changes in the design and the 3D model.

Providing very quick changes to the model potentially allows the undertaking of a deeper study and optimisation of the design.

USE OF PARAMETRIC MODELLING ON THE HS2 EUSTON STATION ENABLING WORKS

In this project, I used the steps described above to implement a parametric modelling design where I created three different scripts in Autodesk Dynamo, one for each main tunnelling asset: the temporary shaft, the ‘long’ tunnel and the ‘short’ tunnel, including the short adit connections. The input database included over 120 geometrical parameters which went to define the whole design.

The temporary shaft was a geometrical challenge. The design included a transition from a segmental lining arrangement to a sprayed concrete lining where the upper and lower centres were offset and so both were connected by means of an asymmetrical flaring. The design of the permanent lining included a cast in-situ complex collar arrangement at the tunnel opening, shown in figure 5.

Changes in tunnel alignment and elevation, tunnel thickness and base slab level had to be made during the detailed design. These changes were due to the changing interfaces, such as the existing asset connections design and with the HS2 Euston Station base slab design. Having the geometry coded allowed getting the changes automatically, and thus provided agility in the redesign.

Building up the script for the alignment of the long tunnel was challenging. This tunnel connected the temporary shaft with the headhouse. Interfaces with the headhouse design and the HS2 Euston station base slab above were the reason why the 3D model of this tunnel had to be updated during the detail design. Applying vectors and tangent compatibility conditions between the straight and curved axis had to be parametrised as well as the cant of the invert.

The design of the short tunnel and connection adits model was a clear example of the main advantage of parametric modelling. It allowed the management of the ‘known-unknowns’ more efficiently – this is the case when key parameters are not yet available.

In this specific case, a geometrical survey of the position of the existing assets had to be carried out while the design was being developed. By coding the geometry before the final inputs were confirmed, I was able to create a progress model which was an estimation of the reality but which could be quickly updated when the survey information became available.

This procedure saved considerable modelling time in the latest stage of the design, reduced the checking time of the 3D model and allowed more flexibility in the design activities, while also reducing the programme.

The Autodesk Dynamo model was exported and reimported into Bentley Microstation – the HS2 BIM software, where additional checks were undertaken to verify the correct transfer. This 3D model was also used as a part of the federated model (Figures 6 and 7) of the project and to generate the detailed drawings.

Using the model I created using the parametric process, the team produced over 60 detailed coordinated design drawings.

Parametric modelling was not only used to generate the 3D model but was also used to optimise the tunnelling section and to link the geometry to the calculations.

At the beginning of the RIBA4 design, an optimisation of tunnel profiles, a redesign of the temporary shaft and the solution for the connection were analysed. By means of a parametric study with the 3D model, a set of alternatives were extracted directly from the model, including their corresponding excavation and concrete volumes. This study therefore included sustainability considerations in the decision making. As a result of this study, the excavation and the concrete volumes were reduced by 10%. These results were presented to support the design changes to the client and HS2, and which were ultimately accepted.

Linking the geometry created by parametric modelling to the calculations was critical to ensure alignment between the 3D geometrical model and finite element analysis built to determine the geotechnical loads on the lining and to undertake the design. Any changes required in the geometrical configuration could be automatically transferred into the finite element package. With parametric modelling, I coded the structural models geometry (Figures 8 and 9). In so doing, I was able to control the geometry to be used in the finite element model as the ‘single source of truth’, ensuring the key geometrical parameters are aligned with the latest version of the design and the FEM was built up as intended.

LESSONS LEARNT: ADVANTAGES AND CHALLENGES

The design of the tunnels and the shaft design is currently completed and submitted for assurance to HS2. Throughout the whole design process, a series of advantages and lessons learnt can be listed showing the benefits of parametric modelling, the limitations found and the potential improvements – plus the further steps that would be required to develop this innovative methodology further.

The main findings and advantages were as follows:

  • The efficiency and quality of the design was increased while reducing the modelling time.
  • Reliability of the design was increased by owning the model, as the engineer controls the geometry throughout the whole design process.
  • A clear audit trail for checking and quality control throughout the process provided assurance to the design by means of the input database and the controlled management of the changes.
  • Tangible savings in design were identified by a decrease in the percentage of CAD time spent compared with other major projects where we were leading the design (Figure 10). An additional saving was realised by using parametric design, even in the delivery of detail design 2D drawings. On the other hand, the cost of additional engineering spending increased. However, for future projects, the engineering cost is likely to be reduced, thanks to the legacy in learning and development that this project provided to the designers.
  • The methodology and knowledge developed has been applied to other projects. I was able to share the findings and advantages of parametric modelling with other teams within Mott MacDonald. I have trained others and helped to implement parametric modelling to suit the design needs. It is not necessary to produce a fully automatised generated geometry, but to adapt and customise the possibilities that parametric modelling offers to maximise the efficiency of the design.

In terms of limitations, a few challenges were found with parametric modelling, such as the production of very detailed parts of the 3D BIM model as well as with the 2D detail drawings. It is important to assess the cost-opportunity of using parametric models against traditional methods (Figure 11). The engineer needs to evaluate whether it is worth coding a specific element or if it is more efficient to exclude that element, and design it using traditional methods. Some elements which are difficult to code can be made very quickly with traditional CAD/BIM design.

There are many other applications that can be applied to the tunnelling design which were identified as potential improvements in the methodology, for example, the possibility that implementing space-proofing rules and constraints improves the coordination with other design disciplines.

Similarly, the use of generative design or ‘refining’ can lead to working out the best solutions for the optimisation criteria provided.

CONCLUSIONS

Parametric modelling is a methodology to code geometry which will be driven by a set of parameters, providing important gains in efficiency compared to the standard design process. This is a further development of what has already been achieved in recent years with BIM technology.

Parametric modelling is not a revolution but an evolution which I have been able to successfully adapt and apply to tunnelling design, specifically to a relevant project in Central London.

Investing in the development of this technique has shown the advantages and provided the lessons learnt to consider parametric modelling as a ‘design as usual’ option and also to keep expanding this methodology into other aspects of the design.

The success of parametric modelling and the potential observed in this pioneer project within Mott MacDonald by successfully delivering a RIBA4 design, constitutes a benchmark for other projects around the globe.

ACKNOWLEDGEMENTS

This paper was possible thanks to the trust the WP137 project team (especially Christoph Eberle) invested in me to carry out this innovative concept, and to the Information Management and Digital Delivery team at Mott MacDonald for their support in taking forward the developed concept and many other innovative initiatives. Thanks also to Rosa Diez for encouraging me to present this paper and for leading the review. And thanks to Sameer Moghal and Bernardo Zornoza for both their review and feedback.