So You Want to Study Machine Learning and Civil Engineering?

Machine Learning (ML) in its literal terms implies, writing algorithms to help Machines learn better than human. ML is an aspect of Artificial Intelligence (AI) that deals with the development of a mathematical model which is fed with training data to identify patterns in that data and produce an output.

In other words, ML is concerned with the utilization of data to train machines (e.g., computer) to learn things (e.g., identifying different
dog breeds) and produce an accurate output when fed with a test data (e.g., this dog is a terrier). The machine gets better (in terms of accuracy of the output) with time and/or with more data. The goal of ML is to make machines that act like humans, to make man’s tasks easier.

Who can study ML?

Anyone with basic understanding and interest in computer science, statistics, and data structure can study Machine Learning.

Brief History

ML is not new! Indisputably, it's the trend today because businesses have realized it prediction potential, so they are throwing in grants and funding to encourage ML researchers to explore it and make findings that will help businesses stay ahead of the competition. On the contrary, Its potential is still yet to be maximized.

Researchers hold somewhat different views about ML start date. Hugo Mayo recorded that the first case of ML was in 1943 when Warren cCulloch
and Walter Pitts decided to create a model of neurons using an electrical
circuit, and it was coined the neural network (NN). On the contrary, in the
words of Dataversity's Keith D. Foote,

"ML model was created in 1949 by Donald Hebb in a book titled The Organization of Behaviour (PDF), which represent Hebb's theories on neuron communication."

The common opinion held by researchers was that Alan Turing created the "Turing Test" to measure computer intelligence in 1950. Subsequently, In 1952, Authur Samuel wrote the first computer learning program and named it "Machine Learning". Since then, researchers such as Frank Rosenblatt, Gerald Dejong, Terry Sejnowski, etc., and tech-inclined companies such as Facebook, Amazon, Google, Microsoft have also been exploring ML till date and progressing.

Application of Machine Learning to Civil Engineering

The application of ML to Civil Engineering began in the 1980s when ML techniques were applied for knowledge extraction from Civil Engineering (CIE) data. The field of civil engineering is rife with the problem of uncertainties in areas not limited to construction management, safety, design, and decision making; the solution to these problems depends on calculations and experience of practitioners.

The procedures to finding the solution to these complex problems are tedious, expensive, and time consuming, as such civil engineers decided to provide solutions to this problems by taking advantage of Machine learning ability to imitate experts.

Kansas State University's Hani Melhem reports that the two most commonly used ML techniques are: Learning from Examples (or Inductive Learning) and Learning from Observation (or Conceptual Clustering). Again, the Hong Kong Polytechnic University's Amos Darko reported that genetic algorithms, neural networks, fuzzy logic, fuzzy sets, and machine learning have been the most widely used AI methods in the Architecture Engineering and Construction (AEC) industry to address pressing topics like Optimization, simulation, uncertainty, project management, and bridges.

Tel Aviv University's Yoram Reich explained that there has not been significant practical progress in using single ML techniques as regular tools by engineers due to the complexity of practical problems, as such ML researchers postulated the use of multiple ML techniques (termed multistrategy learning e.g., Bridger, MOBAL, MLT, and MCS). Further, Reich explained that there have been studies:

• on knowledge extraction (e.g., the feasibility of wind bracing and environmental impact assessment)
• about solving problems in which learning played a major role (e.g., cable-stayed bridge design and monitoring water treatment plants)
• that employed learning as part of their operation (e.g., steel bridge design, highway truckload monitoring, transmission line towers design, and architectural design).
• directed at information modelling to create estimation models (e.g., material stress-strain relations and properties of composite materials)
• on modelling aimed at improving the understanding of a phenomenon (e.g., decision procedures and material corrosion behaviour).

Overall, Civil engineering has only explored a small number of ML techniques, and past studies were limited to little or no standardized
testing and no follow-up work. It was evident that ML was still in its infancy.

Today, ML has experienced some maturity because lots of grants and funding have gone into automation of civil engineering domain among
other fields. For instance, ML researchers have developed:

• Alex-net Neural Network to monitor the structural health
• Back-Propagation Neural Networks to assess slope failure
• Artificial Neural Network (ANN) to predict the compressive strength of concrete.

Similarly, in alignment with the report by the University of Bridgeport's Manu Mitra, Darko reported that Convolutional Neural Networks (CNN)
have only in this modern time been classified and applied as vision and
learning-based methods for solving problems such as damage detection, facility operations and management, construction sites safety monitoring, concrete compressive strength estimation, structural health monitoring (SHM), maximum gradient (MG) and decision-making.

ML researchers are currently looking into the visual inspection of civil infrastructures and collaborative research inspired by the challenging problems faced by visual inspectors in civil, structural and earthquake engineering. This brings into line with Darko's conclusion that there are future research opportunities in using CNNs and robotic automation for solving civil engineering problems.

As a green researcher looking to apply ML to your research, I hope this write up provides you with information about a relevant aspect of ML
to apply to your research. Kindly drop a comment below if you have questions, comments or criticism. Thank you!