EAGCG Recap: IUCM & Abaqus
This year's EAGCG workshop was spectacular. There was an extensive line-up of speakers and sponsors, about 145 registrations. Fantastic venue, and the topic 'Mass Mining' attracted some exciting presentations. Great opportunity to catch up and to network, plus a dinner by the pool. Two thumbs up!
I understand not everyone could attend, so I'll share a recap of my talk in this article: "IUCM for Abaqus and its Benefits for Caving Simulation". Well, it's not guaranteed to be complete or accurate.
So, what is IUCM?
Cavroc's IUCM aims to establish a practical rock mass material model in geotechnical engineering for use in numerical modelling tools like ITASCA Software 's Flac3D or SIMULIA 's Abaqus. By using widely accepted research from well-respected authors, it overcomes the limitations of common models found in commercial software.
Technology
Why?
Why not just use Mohr-Coulomb or elastic behaviour and make conclusions based on stress? The current software landscape in the mining industry still contains many tools and methods that I consider to be outdated. The expectation of having the best geotechnical knowledge and design (and safety!) but using the simplest, lowest-cost software and 200-year-old constitutive models is as compatible as trying to build an A380 with the tools of the early 1900s.
“most engineers do not have access to nor the experience to use high-end modelling software ...“
I've heard this argument. I'm afraid I have to disagree. I don't want the university book club designing the plane I travel on just so everyone can have a go. I do want experts, experience and the best tools for the job. We should expect that, in the industry that in Australia contributes more to GDP and exports and jobs than any other (AusIMM, Reserve Bank of Australia), we can reasonably expect to set the highest standards. Yes? Well, let's move on ...
Caving Geomechanics
There are lots of conference papers and models and discussions about caving, but one had me intrigued back in 2018: The observed behaviour of centrifuge models collides with model results (all links in the images are under 'References')
I also thought examining models that don't work would be illustrative. Everyone likes to share their best calibration, so let's look at the bloopers. Now, the videos don't work here, see youtube. In defence of the Flac3D model on the right, I think the chimneying is into the overlying panel, so that's already broken. Still ambitious to pull that column. On the left is an Abaqus model that was shown on a few occasions, and the fact that a trial set of parameters here leads to chimneying through intact rock makes it clear this is not reality. The problem here for model calibration was using Mohr-Coulomb with strain softening.
That is not sufficient. Rock mass has a very complex confinement-dependent peak-residual stress-strain relationship. Mohr-Coulomb can not produce results like Itasca's IMASS or IUCM. Strain softening and dilation have an important role in the evolution of damage, as demonstrated in the tunnel example for IUCM below that replicates the observed failure.
The Elephant in the Room
This brings us to the elephant in the room, borrowed from Davide Elmo 's Rock Engineering Aphorisms.
“Is the concept of rock mass strength universal or relative to the modelling approach used in the analysis?”
You would hope that there can be clear guidelines on determining rock mass parameters applicable to numerical modelling (you'll find these in the cavroc forum for IUCM). The choice of software should not be based on which results you'd expect to get.
Flac3D, Abaqus ... it is all the same
I've been saying this for a long time. Finite Element ... Finite Difference ... as long as you solve non-linear equations with a comparable material model, you should get the same results. It's all the same (also discussed in the recent article on Slope Stability, Monkeys and Bananas). ITASCA Software Flac3D is an excellent tool. SIMULIA Abaqus is great too. That's it. Sorry to disappoint anyone who was expecting a revelation here.
Exploding Kittens
Then there was the anisotropy explainer - I'll refer to How rock anisotropy works in numerical models
Travelling at the speed of light
I'm biased, as I have been using Abaqus for 30 years. I think the performance is great (I don't think, I know), and it scales very well for parallel performance (and further, you can watch a benchmark car crash and webinar replay mentioning ~1800 CPUs). I know it is considered the expensive option. That's for the solver.
What I consider a great success in our IUCM implementation in Abaqus is speed. I don't want to compare Apples and Bananas, so the benchmark here is the Abaqus built-in elastic model (that's basically six lines of code for stress-strain from Young's Modulus and Poisson's Ratio). We achieve over 30% of the speed of an elastic model with IUCM.
Another advantage of using Abaqus is the fact that there is an implicit solver. If your stress can not be described with an analytical formula using a linear gradient of rho * g / meter with the principal orientations horizontal and vertical, then achieving equilibrium can be time-consuming in explicit solvers. Solving this equilibrium stress state (from which the actual analysis continues) can be done directly with implicit solvers (either direct sparse or with the iterative solver in Abaqus that is fast, requiring less memory for large models). This is very useful.
领英推荐
That brings us back to the comparison. For this example, the comparison with the Flac3D version is less accurate because of time and speed constraints (apologies to the Itasca team, it's us, not you!), but you would still draw the same conclusions and have the same forecast for overbreak.
What you see is what you get
A black box approach isn't ideal for achieving consensus among numerical modelling experts, software users and decision-makers. IUCM is transparent, based on the publication by Abouzar Vakili (2016). Showing a few lines of our C++ code for Abaqus VUMAT gives you an idea.
At this point, it is worthwhile to mention AI taking over soon (see Monkeys on Typewriters ... what is it with this guy and the monkeys? ...) - so the above code is not a differentiator for skills anymore but could just about come from Copilot.
IUCM Features
IUCM has been available for Flac3D for many years, with a wide user base and integration of StopeX and SlopeX. It is now available as a constitutive model library for Abaqus/Explicit and on the cloud in the 3DEXPERIENCE platform.
Caving Simulation
This brings us to the application for caving. For this simple toy model of a block cave, the material model can replicate observations in the physical models.
This concludes the discussion (I had to skip parts here – this is just a recap). I would like to point to German Flores' excellent keynote at the ACG Australian Centre for Geomechanics Mining Geomechanical Risk conference 2019, providing the industry with a solid discussion plus excellent illustrations. The key point here is the impact geomechanics can have on the production forecast (even with all the 2022 model limitations mentioned above).
No one looks smart with VR goggles on their head
One more thought on the complex 3D results that we obtain through simulation (and the cost and hours that go into creating them): Is it time to put the goggles on? My point for many years has remained strong. There will be a difference between the decision-making process of looking at a risk heat map or probing into a 3D virtual twin to gain deeper insight.
Summary
We keep looking to other industries to understand where we want to see ourselves in the shortest practical timeframe when it comes to technology. It's time to let go of simple tools and bad habits.
We do hope the discussion around the IUCM model can promote ideas regarding more standardised numerical modelling and consensus in the geotechnical engineering community.
So, what constitutive models are you using and, importantly, why?
References
https://doi.org/10.36487/ACG_repo/2205_64 (Arndt 2022)
https://dx.doi.org/10.1016/j.compgeo.2016.08.020 (Vakili 2016)