Machine learning is helping us find new answers to old problems, and outcomes for fresh challenges presented by our ever-evolving world.

Machine learning is a type of artificial intelligence in the field of computer science, that gives computers the ability to ‘learn’ repetitively from data through algorithms. It can be applied in a range of fields and industries, but use is increasing across engineering, design and infrastructure. 

Machine learning can help to generate, manage and make sense of data, providing meaningful insights to assist Aurecon and our clients make better decisions for planning, building or optimising infrastructure. For example, machine learning can be used to help utility organisations predict daily demand for energy consumption when planning residential developments.

Machine learning operates closely with other digital technologies such as big data, analytics, data management, visualisation, augmented reality and virtual reality as we create smart infrastructure for the future.

Machine learning opportunities in engineering and infrastructure

Previously, the limits of computers, process complexity and access to relevant data of sufficient quality limited progress.

However, advances in technology along with increases in data generation presents new opportunities to make sense of complex systems. The advent of data to plan and build infrastructure, then operate, maintain and monitor it with unsurpassed quality of information is opening many doors.

Machine learning can be used in a variety of ways across engineering and infrastructure, including modelling traffic systems, streamlining manufacturing processes, predicting energy demand, forecasting equipment maintenance, simplifying building and project management and communications.

Aurecon’s Machine Learning Centre

Our Machine Learning Centre brings together the brightest minds in science, technology, management and mathematics across academia and industry, collaborating with central engineering teams in supporting their design and consulting work.

We collaborate to apply machine learning tools, platforms and techniques to address client needs, deliver improvements and new insights by combining our expertise with emerging digital capabilities.

Machine learning applications 

• Data analysis 
  Data analysis using statistical and pattern recognition techniques to provide greater understanding of data relationships or system process can inform design or operational decisions. Aurecon has applied this approach involving machine learning techniques to undertake capacity assessment and root cause analysis for model mechanical plant operations and process engineering.
  
  
• Adaptive learning
  Instead of relying heavily on assumptions of data fixed in time, a system model makes use of new or continuous data to provide more accurate predictions and efficient operations of infrastructure. We developed a machine learning based tool to automatically conduct pipe inspections, removing the need for manual inspection. Our tool reduces the need for time-consuming visual inspections that traditionally place large burdens on personnel with fatigue adding to the risk of errors, aiding engineers in their operational process and quality assurance requirements.
  
  
• Feature extraction
  We have used machine learning with an unmanned aerial vehicle (UAV) photogrammetry tool to complete digital rock mass mapping more accurately, quickly, safely and remotely on treacherous mountains in South Africa and New Zealand. Machine learning based feature extraction has also been applied in our road surface evaluation work.
  
  
• Event and data detection and classification
  Anomaly detection can be performed considering past history and patterns of ‘normal operation’ to detect variations from the norm, as well as classifying past events of interest. We have applied this in developing automated incident detection capabilities for traffic monitoring.
  
  
• Predictive modelling
  This enables forecasting or determining the likelihood of events. For example, at Aurecon we have used historical metering data from a residential development to predict critical peak loads for energy usage to enable a utilities organisation to plan infrastructure for a new development. 
  
  
• System optimisation
  Where reliability and accuracy of system optimisation is sufficiently proven, machine learning outputs can be incorporated into system controls to better leverage more in-depth data analysis. This applies to HVAC optimisation in configuring system operations to leverage modelling outcomes.