Very interesting. I recently conducted a brief review of Applications of AI in Geotechnical Engineering and i could hardly find any case studies. So, yes in that sense, the application of ML to Geotechnical problems is still an academic debate. However, the way it is gaining the currency in our time, it is hoped that the academic knowledge regarding AI is taken to the field very soon. After all, every new thing has a gestation period !

The application of AI in the field of Geotechnical Engineering is highly dynamic. One can easily imply various machine learning model to determine the desired result that requires extensive human involvement and time consumption. Ali Parsa-Pajouh I am recently working on the same, you can have a look on my publications.

If you need any regarding the same you can contact me.

Dear Ali Parsa-Pajouh

This special issue might be useful, have a look

https://www.mdpi.com/journal/geosciences/special_issues/applications_artificial_intelligence

Kind Regards

Qamar Ul Islam

Thanks Qamar Ul Islam for sharing the link.

Thanks Sufyan Ghani for sharing your works.

It seems that liquefaction prediction is one of the areas that has great potential for application of AI and ML.

I also found the following publication about the use of machine learning to predict lateral spreading.

Article An exploration of the use of machine learning to predict lat...

Your very welcome Ali Parsa-Pajouh

Stay Happy Stay Healthy

Here are some of the highly cited articles about application of AI and ML in geotechnics.

• Machine learning in geosciences and remote sensing.

  Article Machine learning in geosciences and remote sensing

• The application of genetic algorithms to optimisation problems in geotechnics.

  Article Application of genetic algorithms to optimization problems i...

• Metaheuristic optimization within machine learning-based classification system for early warnings related to geotechnical problems.

  Article Metaheuristic optimization within machine learning-based cla...

• Prediction of Geotechnical Parameters Using Machine Learning Techniques.

  Article Prediction of Geotechnical Parameters Using Machine Learning...

• Machine Learning Techniques in Landslide Susceptibility Mapping: A Survey and a Case Study.

  Chapter Machine Learning Techniques in Landslide Susceptibility Mapp...

• Landslide identification using machine learning.

  Article Landslide identification using machine learning

• Prediction of soil compression coefficient for urban housing project using novel integration machine learning approach of swarm intelligence and Multi-layer Perceptron Neural Network.

  Article Prediction of Soil Compression Coefficient for Urban Housing...

• Machine learning in soil classification.

  Article Machine Learning in Soil Classification

• Prediction of lining response for twin tunnels constructed in anisotropic clay using machine learning techniques.

  Article Prediction of lining response for twin tunnels constructed i...

• Machine learning for pore-water pressure time-series prediction: Application of recurrent neural networks.

  Article Machine learning for pore-water pressure time-series predict...

• Estimation of Bearing Capacity of Piles in Cohesionless Soil Using Optimised Machine Learning Approaches.

  Article Estimation of Bearing Capacity of Piles in Cohesionless Soil...

• Application of artificial intelligence in the field of geotechnics and engineering education.

  Article Application of artificial intelligence in the field of geote...

• Recent Advances and Future Challenges for Artificial Neural Systems in Geotechnical Engineering Applications.

  Article Recent Advances and Future Challenges for Artificial Neural ...

• Prediction of shield tunnelling-induced ground settlement using machine learning techniques.

  Article Prediction of shield tunneling-induced ground settlement usi...

• A machine learning approach to energy pile design.

  Article A machine learning approach to energy pile design

• Risk assessment and management of excavation system based on fuzzy set theory and machine learning methods.

  Article Risk assessment and management of excavation system based on...

• Analysis of diaphragm wall deflection induced by excavation based on machine learning.

  Article Analysis of Diaphragm Wall Deflection Induced by Excavation ...

• Application of deep learning algorithms in geotechnical engineering: a short critical review.

  Article Application of deep learning algorithms in geotechnical engi...

• Application of artificial intelligence to rock mechanics: An overview

  Article Application of artificial intelligence to rock mechanics: An overview

• Prediction of Swelling Pressure of Soil using Artificial intelligence techniques

  Article Prediction of Swelling Pressure of Soil using Artificial int...

Interesting discussion and great work Ali Parsa-Pajouh. I believe major hinderance in Geotechnical engineering regarding the use of AI is data availability (amount of data) and scale issues, specially problems like estimation of bearing capacity, settlement analysis, strength of subgrade etc. On the other hand, numerical analysis techniques give us the advantage of generating useful data if we can validate some large-scale experiment (e.g., paper quoted above in Computer and Geotechnics). Moreover, there are no standards developed for systematic application of these AI/ML techniques in Geotechnical Engineering. But this issue is same with other areas (structures, hydro, transport etc.).

Another main concern is the easiness of use for the developed AI models. As a practitioners, one would like to simply plug and play the values and get the answer . For this, if we as researchers can convert our AI models into mathematical equations for hand or spread sheet calculations would be great.

Have a look at here too Article Application of Machine Learning in Geotechnics

and https://www.booktopia.com.au/applications-of-computational-mechanics-in-geotechnical-engineering-r-f-azevedo/book/9789054108641.html?source=pla&zsrc=go-nz-allstock&gclid=CjwKCAjwrPCGBhALEiwAUl9X03v3y-6E1Pq5Zcm01S2OUZQFxCSNwU3fD5SpqhvfIxVrNQHL7gKpwBoCtfUQAvD_BwE

Muhammad Nouman Amjad Raja Thanks for your helpful contribution to the discussion.

That's right, one of the major challenges in this area is the lack of reliable data in large scale. However, from my industry engagement experience, I can say that the data is available but the main issue is the lack of effective collaboration/communicate between universities and industry to use these value databases for AI/ML studies/analyses, which by itself could be another important topic for discussion.

I believe the Geotech industry should be well educated through simplified manuals and design methods to learn how to use AI and ML approaches in practice and also be able to review/verify the designs that have been done based on these techniques.

Applications of AI on Geotechnical Engineering can be wide, especially in geotechnical investigations but scarcity of data makes it difficult. There is this article from Pr Ebid which can be of interest: 35 Years of (AI) in Geotechnical Engineering: State of the Art

https://link.springer.com/article/10.1007/s10706-020-01536-7

Interesting discussion, there are several axis of research for the application of AI and ML in geotechnics such as prediction of settlements and landslides, soil properties, soil equifcation ... etc. but the problem that poses every day is the lack of data to make a good model.

Dear Ali Parsa-Pajouh ,

In geotechnical engineering, the investigation of various geological hazards such as landslides is one of interesting and important research topics.

The use of ML in the investigation of geohazards is interesting.

Quite recently, our group published a review paper on

"Machine Learning for Landslides Prevention".

It might be useful, please have a look at:

Article Machine Learning for Landslides Prevention: A Survey

Kind Regards

Gang Mei

Dear Gang Mei ,

Thanks for your comment and sharing your very interesting and technical paper.

Kind Regards,

Ali

Dear Ali Parsa-Pajouh,

There are many practical applications, for example, models for detecting microcracks in hydraulic structures using (vibro) ultrasonic signals. ML& ANN models are most often used in signal processing methods.

Dear Ali Parsa-Pajouh,

Artificial intelligence methods (machine learning and deep learning algorithms) are now well established in geotechnical engineering, rock mechanics, tunnelling construction, and soil mechanics, and can save both time and money. Machine learning is a powerful technique to predict the target parameters with high accuracy and precise during pre-construction, construction, and post-construction phases. It would be helpful, if it would be possible for you to take a look at the following papers:

Brian B. Sheil et al., 2020. Machine learning to inform tunnelling operations: recent advances and future trends.

Thomas Marcher et al., 2020. Machine Learning in tunnelling - Capabilities and challenges

Samadi and Farrokh, 2021. Utilization of Rock Mass Parameters for Performance Prediction of Rock TBMs Using Machine Learning Algorithms.

Best Regards,

Dear Ali Parsa-Pajouh,

What I can imagine is that with the expansion and innovation of AI/machine learning, the application of AI/machine learning in civil/geotechnical/geological/mining will definitely be all over the place.

I would like to give an example, in geotechnics, experimental/empirical formulas in textbooks, reference parameters in engineering manuals, etc. have been widely used all over the world for the past decades or even centuries. However, there is a gap between these empirical results and the real world, and this gap may be able to be rechecked by means of AI/machine learning, etc., and present more extensive, accurate results.

On the other hand, there is a large amount of data accumulated in engineering practice, and people are often ignoring the hidden connections and patterns between or within the data due to the lack of proper analysis methods or computational power before, so why not use mathematics and statistics based AI to explore it? This should be very interesting.

But I would say that methods are methods after all, and they are brought to bear only to solve real-world problems. And today, young scientists seem to be chasing fancy AI methods, ignoring how simple and effective they can be to solve what kind of engineering problems.

Best regards,

Chengkai

I attended TC304-TC309 Pre-conference Workshop on 'Risk Assessment and Machine Learning in Geotechnical Engineering' during the ICSMGE 2022 Conference in Sydney on 2nd May 2022 chaired by Prof. Jinsong Huang of Newcastle University and Prof. Mohammad Rezania from University of Warwick. It was a vey informative session with attendance of over 200 online and in-person experts from around the world. (https://lnkd.in/em4v6hgJ)

Please check the link below for the recording of the session with all state-of the-art presentations:

https://lnkd.in/em4v6hgJ

Gang Mei Chengkai Fan Dunshi Wu

Thank you very much Ali Parsa-Pajouh for sharing the link. Very interesting !

Dear Ali Parsa-Pajouh,

In the last biennium, several journals have published special issues on Machine Learning applied to geotechnics. Below are three references:

Georisk: https://www.tandfonline.com/toc/ngrk20/16/1?nav=tocList

Acta Geotechnica: https://link.springer.com/journal/11440/volumes-and-issues/17-4

Risk and uncertainty in eng systems-ASCE: https://ascelibrary.org/ajrua6/bayesian_methods_geotechnical

Best regards,

ÁJ Mattos

"Machine Learning in Geotechnical Engineering", an ISSMGE Recorded Webinar

Presenter: Prof. Pijush Samui

(https://www.issmge.org/education/recorded-webinars/machine-learning-in-geotechnical-engineering)

Geotechnical instruments collect data, and when there is data AI and ML can very well be applied. Well, while talking about practical applications it could be any of the following

(1) Finding the relation between two parameters. Example: Rainfall and Moisture

(2) Creating AI and Ml models for predicting, forecasting and early warning. Example: With time series data of particular geotechnical sensor, understanding its relationship to heterogeneous triggering factors from the historic data and trying to forecast the futuristic conditions.

(3) Developing thresholds

(4) Mapping the unknown relationship

(5) Modelling the natural phenomenon

Internet of Things (IoT) is another promising field where the recent trends are being established in geotechnical engineering

Inclinometer and tiltmeters with industrial IoT capabilities, automated data logging, and visualization are readily available in the market. Industrial IoT piezometer sensors are already employed for monitoring structural health in bridges. Industrial IoT geotechnical sensors are already deployed in place for water monitoring systems, air quality monitoring systems, environmental monitoring, and deploying such sensors for slope monitoring applications can act as IoT EDGE device.

I have dedicated a section for IoT and Geotechnical Engineering in my recent paper “Review of Landslide Monitoring Techniques with IoT Integration Opportunities”.

If interested please have a look at it. Article Review of Landslide Monitoring Techniques With IoT Integrati...

Ali Parsa-Pajouh Thanks for the interesting question.

Dear Ali Parsa-Pajouh

within the framework of strain-dependent slope stability for earthquake loading, in a recent publication with Jan Machaček, Luis Felipe Prada Sarmiento, Patrick Staubach, and Torsten Wichtmann in Computers and Geotechnics, we used ML to replace element test simulations performed with advanced constitutive models. Machine learning algorithms adapted to the problem of a water-saturated slope subjected to earthquake loading provided accurate approximations of the factor of safety and the failure mechanism while reducing computational costs by 2-3 orders of magnitude.

Article Strain-dependent slope stability for earthquake loading

Ali Parsa-Pajouh

AI and ML techniques are really helpful for liquefaction potential assessment of soils. Check this latest article for more details: Article Explainable machine learning model for liquefaction potentia...

We have done some work in the space. https://www.researchgate.net/publication/366091050_Calibration_of_numerical_models_to_improve_tailings_management_through_Bayesian_methods

The biggest difficulty in AI is to provide a large data for the problem.

This paper Article CPT-based method using hybrid artificial neural network and ...

Artificial intelligence (AI) and machine learning (ML) techniques are increasingly being used in geotechnical engineering for a variety of purposes. Some examples of practical applications of AI and ML in geotechnical engineering include:

• Predictive modeling of soil behavior: AI and ML techniques can be used to build predictive models of soil behavior based on data from laboratory tests and field measurements. These models can be used to predict the behavior of soil under different loading and environmental conditions, which can be useful for designing geotechnical structures such as foundations and slopes.
• Geotechnical data analysis: AI and ML techniques can be used to analyze large datasets of geotechnical data, such as soil data from drilling logs and geotechnical instrumentation data. This can help engineers to identify patterns and trends in the data, which can be used to improve understanding of soil behavior and inform design decisions.
• Automated soil classification: AI and ML techniques can be used to classify soil samples automatically based on data from laboratory tests and field measurements. This can help to reduce the time and effort required to classify soil samples manually, and can also improve the accuracy of soil classification.
• Geotechnical risk assessment: AI and ML techniques can be used to assess the risk of geotechnical failures, such as landslides and slope failures. This can help to identify areas that are at high risk of failure, and can inform decision-making about how to mitigate these risks.

It is not uncommon for AI and ML techniques to be used in academic studies in geotechnical engineering, but they are also being used more and more in practical applications. There are currently no widely accepted standards for the systematic application of AI and ML techniques in geotechnical engineering, but there are ongoing efforts to develop guidelines and best practices for their use in this field.

I think it is all about making coreelation and the consideration of the uncertainty in the model as well as in the model data. In rhis regard, it may sound like nothing more than statistics; however, there is quite a lot from this, especially for handling highly complex, non-linear, large data involving, and ambigous mappings between input and output spaces that are in question. On the other hand, similar to the statistics, the performance of the model or the system is highly related with the quantity and the quality of the data, thus, it make sense to consider the results or inference together with certain relaibilty levels and/or the strength of correlation with appropriate measure of closeness. it must also be noted that the selection of free variables (or model parameters) is of particular importance as in statistics.

Under the light of these, there is a plenty of application areas of AI or ML in geotechnical engineering. At the end of the day, making coreelation through neural netorks or deep learning, handling data or making inference with the help fuzzy logic for the problems with considerable ambiguity or vaugeness, or finding the values of free parameters to maximize (or minimize) an objctive function via genetic algorithns or swarm intelligence, finding out the boundaries of distinguishment for different behviors (namely, classification or understanding belongingness to a certain group) with a measure of closeness using self organizing maps, fuzzy logic, or etc., or making inference using linguistic variables or establishing correlations with predicate logic as in expert systems, or making correlations and identifying the valuss represnting a behvior using intelligent agents, or identification of a pattern or making a forecast after realizing the relationship, or extrapolation or classification algorthms based on AI methods, are nothing more than the establiablishment of a llinear/nonlinear mappings between input or output spaces, which is the basic defintion of finctional analysis in mathematics.

In my opinion, more important than what the problem is, it is crucial to select appropriate model paramters, to understand the uncertainty in the system, to have as much as possible good quality and enough quantity (and distribution) of the data, to interpret the resilts logically and properly, and most inportant of all to select appropriate approach/method for the problem which is either AI or staitstical based. It should also be noted that in seveeral cases in geotechnical engineering, the uncertainty in model parameters or data in hand is quite a lot (in most cases, there may even be lack of appropriate or sufficient data) and there are certain behaviors that are expected to be observed. Therefore, the precission of the model (for examples in terms of R2 value), may not be good indicator of the success since the ability of learning (or mapping) highly complex behavior via AI techniques may include the learning of error or inprecission itself in the data in hand, which may result in misleading conclusions theough the model developed.

I truly beleive and experienced that it still matters the most who is developing the model, selecting the parameters, and collectong the data with proper test method or observations, are still the main challenge to select the best fit for the solution/modeling as well as inference methodology. We should never forget that the uncertainties or errors or missing knowledge in the data we have or the correctness/appropriateness of the model we use for the represnting of soil/rock behaviors may be much more important than the increase in the closeness or success ratio of the mapping.

As a resercher and an engineer spent most his carrier on geotechnics, foundation engineering, statistics, and AI, I would like to underline that data, selection of proper and "required" methodology,undetanding the problem and behvior correctly, and selecting suitable inference method are more important than the decision on the selection of AI based or traditional approach.

In this regard, AI methods such as deep learning or ANNs are very successful universal finctional approximators and decinitely have wide range of areas of applications in geotechnical engineering; neverthless, they are not methods altering numerical techniques such as FEM and FDM, or constititive soil models which having certain mechanical basis as well as mdel related parameters. I expect, me hanical models and AI techniques wil join the forces in the near future, and there will be certain advantages of doing this.

Moreover, in oder us to observe more AI or ML techniques in geotechnical engineering, it is also needed to comprise them in some software packages and remove the "black box" related hesitatilons.

I can list here numerous application areas or opnions for using AI or ML technigues in GE; however, I beleive it is more important to understand the philosphy behind and to highlight the pitfalls for AI/ML based modeling attempts.

Ali Parsa-Pajouh AI and ML can be applied in liquefaction potential assessment. Still, some significant development is going on by the researchers. You can see my recent review article in this regard : Article Liquefaction Potential Assessment of Soils Using Machine Lea...

Recently, Sydney Chapter of Australian Geomechanics Society (AGS) organised a very informative round-table discussion on ‘ChatGPT and Geotechnical Engineering Practice: Navigating Opportunities and Challenges in the AI Era’ via Clubhouse. Further details about the session and panellists can be found in the attached file.

Replay of the discussion can be accessed via the following link on AGS website:

https://australiangeomechanics.org/downloads/geo-conversations/