Article provided by Neil Chadwick, Director, Digital Geotechnical and Stephen Lawrence West, Director, Ground Engineering, Ramboll

Digital transformation is a term that, within the ground engineering sector, can engender feelings of hope or fear or possibly both at the same time. The hope is for a future where we have tools that can help us all to do a better job for society’s benefit. The fear is of being left behind, individually or corporately, or concern about potential undesirable consequences if we allow the machines to take over. In this article we will explore some of these fears, a few of which are well founded, but our main goal is to accentuate the positives that digital transformation can bring and encourage creative thought and debate about this issue by the AGS members.

It starts with the data

The digital revolution is also a data revolution. The good news is that our industry is already pretty good at handling data. At the heart of this is our own AGS data transfer format for factual GI data. We take it for granted but it is probably the world’s most successful ground data transfer format. Other countries, and other parts of the construction industry, look at us with envy.

Despite this success we should not rest on our laurels – and we are not. The AGS format is adapting and expanding. We have introduced AGSi, for ground models and interpreted data, and a draft for AGS piling is now starting to gather some real interest. The AGS Instrumentation and Monitoring Working Group is also on the case, currently studying real world I&M data flows.

Few would doubt the value of digital tools for helping us to sort out the data, or the need for common standards for when we need share that data. The opportunity that digital transformation provides is a step change improvement on what we already have Our day to day experience of modern apps and websites has, quite rightly, increased expectations for the user experience of technical software. We want tools that are both smarter and easier to use. However, if we want our software vendors to provide these, then we need to be pro-active in telling them what we really need, and why.

Interpretation: humans vs machines?

Interpretation of data to inform design is perhaps the first flashpoint in the discussion about the nature and extent of machine input to our processes. We all know and accept that ground data is normally less than perfect, with outliers that are outliers for a good reason, albeit we don’t always know the reason at the time. In this country we have been reluctant to use even relatively straightforward statistical methods in our interpretation, so therefore it comes as no surprise to find that many get nervous when the digital evangelists start talking about data driven design and machine learning.

It is right that we should question these. Data driven design is just that. In many cases, a data driven design could give us the right answer, but data is subject to imperfections, bias and limitations. If we are not careful we may end up reinforcing the bias (as Amazon once found out, to their embarrassment), or we could get things completely wrong if we extrapolate outside the valid range of the data.  We also need to ask what is the ‘right answer’ when considering design solutions.  Can a purely data led design provide an answer that is right for the overall needs of a particular client and project that suitably weighs risk.  This is where the human element can provide insight to select the right answer having been guided by information provided from past data.

Machine learning goes further and can potentially unlock more value from our data, but the mantra here is ‘don’t forget the physics’. The realities of the ground are such that the machines will always need a human partner for their learning process, whose role will be to  define the geological and other rules that should be obeyed, and act as final arbiter to select the design ground model.

An example of the above is interpretation of geological horizons from borehole data. We already use computers to help us with this, but at present it typically needs need human intervention to account for features such as buried river channels that may be apparent from the desk study conceptual model, but may be have been missed by the existing boreholes.

Having said all that, the authors believe that we should be embracing these techniques, using them to help us make more informed and hopefully better decisions. The fact that 100 different engineers can come up with 100 different design lines from the same data is not something we should be proud of.

A good example of a positive experience from machine learning was given at the AGS Data Conference in 2017. One of the presentations looked at the machine learning applied to CPT interpretation for a large regional scale dataset. Human verification of selected interpretation was undertaken for control purposes. It was confirmed that some of the machine generated interpretations required correction after review, but these were outnumbered by the number of cases where, after review, the machine was considered to have got it right, not the original human!

Automation of the design process

This is where things start to get really interesting. There is plenty of scope for using automation to help with factual and interpretative reporting, but automation of analysis and design calculations is likely to be one of the main digital battlegrounds in the coming years.

As mentioned above, digital 3D ground models are already a reality on many projects. One area ripe for development is finding a better way to input these models directly into analysis/design software, replacing what is currently mainly a manual process. However, we need to think carefully about what model we use for input. We are most familiar with the ‘geological (observational) model’ which is our best guess of what we think might be going on based on the conceptual model and lessons learned from our education and experience. However, for analysis we should be using a ‘geotechnical design model’ which also takes account of uncertainties and code requirements.  These models can be compiled by an intelligent digital partner by prompting the geotechnical engineer to make key decisions relating to geological setting and how to account for ground related risks.

One of the main concerns expressed about automation of design is that we will forget how to do the design calculations, with the younger generation not learning at all. This is a legitimate concern, and it is by no means unique to ground engineering. If we automate (or perhaps when we automate is more correct) we will need redefine the role of the human in the design process. We will probably still be doing some sort of verification, which should allow us to practice our skills and judgement. We don’t have all the answers to this conundrum, but it would be wrong to allow this to be a barrier to much needed progress. We will have to work through it.

There are some who are ambivalent, or even hostile, to automation, fearing that many will lose their jobs. However, others see the opportunities that it can bring, such as allowing us to do more analysis, considering more scenarios, to create better designs for our clients. If we get this right, we should be able to spend more time on real design and less on manually transferring data from one bit of software to another. We need to make sure that, as an industry, we are all aligned in looking to deliver real change and real benefits. We must avoid being drawn into a race to the bottom (on time/cost). It is within our gift.

Automation in construction

Digital transformation is not confined to the design office. There are many opportunities for increased use of digital technologies on site.

Digital field capture of data for ground investigation and construction is now starting to become routine, although we still have some way to go on this.

Augmented reality, where models and the real world can be visualised together, is an under-used technology that certainly merits further attention.

Automation of construction processes is the next frontier. Will we see robots running around construction sites?  One day, perhaps, but the reality of construction automation may be slightly more down to earth. For example, it may include automated boring or piling rigs, or earthworks equipment, with humans still in attendance but with more of the work and decision making done by the machines.

Automation of earthworks operations is the one of the main subjects of research and development in the ground domain, and there are already many examples of digital technologies being used to good effect. Typical applications include tracking of compaction plant, to provide information on number of passes, to assess specification compliance. Another example, published recently in NCE, shows how a contractor is monitoring earthworks vehicle movements on a large linear infrastructure site, then using AI to optimise utilisation of the fleet.

This is not digitisation for the sake of digitisation as there are some important additional benefits that can be obtained by automating construction. Firstly, there is the obvious benefit to health and safety if we can keep as many humans as possible out of harm’s way. This is one of the main drivers for research in this domain.

The other benefits are perhaps not so obvious, but should be of interest to ground specialists. Digitisation and automation of construction will very likely lead to increased monitoring of the processes. This additional data could prove very useful, if we choose to leverage it. A further benefit should be improved consistency of processes which, when taken together with the extra monitoring, should leave us with better build quality, and better records.

We will still need humans to keep a close eye on things as the ground never ceases to conjure up surprises, but if we get this right we could end up in a much better and safer place.

Do we all need to be digital experts?

A good question, that has been answered many times by many people, with many different answers given. The reality is that we can only hold a limited amount of knowledge in our heads. It is unrealistic to ask all ground specialists to become digital specialists, and the authors believe that it would be wrong to significantly dilute or reduce most individual’s knowledge of ground engineering (whatever branch that happens to be) to make way for lots of digital skills.

However, getting digital specialists to do all of our digital legwork is not the right answer either. We should aim for a general increase in awareness and knowledge of digital issues and capabilities at all grades, including the oldies. Digital specialists may still be brought in for the heavy development work, but it would also be helpful to have some people with a foot in both camps.

Whatever the arrangement, we need ground and digital specialists to work together to identify the problems that need to be solved, and the improvements that can be made. If we leave it all to the digital people then we may get lots of shiny new toys, but not the ones we really wanted.

In conclusion

Much of this article has talked about some of the potential problems that digital transformation may bring. However, our mission has been to inform, challenge, and hopefully allay some fears. Digitisation is coming, and we should embrace it, as it will offer great opportunities for improvement within our industry and can be seen as an aid to help us communicate even more effectively with our colleagues, clients, and the public.

However, if we are going to get the most from it, we need ground specialists to work with the digital specialists. The digital specialists may know how to get there, but we need to tell them where to go.