Areas machine learning is utilized in the field of civil engineering research.

The quest for every scientific endeavour is to collect data and make sense out of it. Engineering was born from understanding the behaviour of systems and materials through analyzing data and mapping patterns from it.

Civil engineering is a branch of engineering, with its principles always relying on empirical methods, where data collection is paramount for understanding materials behaviour and predicting structural response.

With the recent advancement in Artificial intelligence and machine learning, where computer systems can understand patterns from data, interest from researchers in civil engineering is growing exponentially.

Some of the areas researchers are actively utilizing machine learning in civil engineering are as follows.

1. construction management:

For every civil engineering project to be successfully executed, cost, quality and time must be the governing direction that all stakeholders will follow. Cost is the most significant factor for deciding project success. Many researchers are utilizing artificial neural networks to come up with models that will predict the cost of capital intensive projects like bridges, highways, sports fields and cement plants at the initial phase of projects when design data is not available. These models are trained using past projects data, using design parameters as the features and the cost as the label.

2. Structural engineering:

The efficiency and precision of how some things in nature carry their weight and distribute it to the ground have been a concern for structural engineers for centuries. Many iconic structural designs we see today were inspired by nature. Examples include the Beijing national stadium, inspired by a birds nest. Structural engineers are now able to design better structures thanks to generative design. Machine-learning has also been applied in structural health monitoring where large vibration data are analysed by either a support vector machine or a neural network and classify whether a structure is damaged or not.

3. Geotechnical engineering

Since most civil engineering structures are anchored to the ground, the issue of understanding the behaviour of soil under different stress conditions has been the interest of a subfield in civil engineering, called geotechnical engineering.

Understanding soil behaviour under different stress conditions requires an engineer to collect huge soil samples, conduct laboratory experiment and analyzes the data, which is on most occasions, time labour intensive and time-consuming.

Machine learning is becoming more applied in areas of geotechnics like the prediction of optimum moisture content and maximum dry density from using simple index properties like liquid limit, plastic limit, shrinkage and particle size distribution. Such models are designed using artificial neural networks and they have the potential of reducing labour intensive and time-consuming experiments like proctor compaction tests.

4. water resources and environmental engineering:

There is a saying that water is life. That means in its absence comes the end of all life that exists on earth.

Despite water sustaining life to most of the earth species, it has also a dark side of creating destruction through events like flooding.

Machine learning applications to water resources have increased in recent years, allowing researchers to offer novel solutions to challenging problems with excess rainfall which induces flooding.

Many researchers have reported promising outcomes of deep learning models on radar rainfall nowcasting or high-resolution forecasting of precipitation. The use of deep learning models (ConvLSTM) in radar rainfall nowcasting and flood forecasting has been studied. The deep learning model employs a blend of convolutional neural networks, which are extensively used in computer vision tasks like face recognition and image classification, and recurrent neural networks, which are commonly used in language translation. To anticipate future water levels at 5 places in Singapore's Bedok Catchment, deep learning models were trained using past observed radar data from Meteorological Service Singapore. The result of such studies could be beneficial to tropical countries with changing weather patterns.