Pilot project – replacing inspection checklists with AI solutions

Inspection² were invited to work on an onshore pilot project for a Global Oil & Gas company. Being keen to support the community and protect the environment, the customer wanted solutions to reduce cost, improve well productivity and recovery, and manage the environmental and social impacts of production.

Naturally, they thought of us with our renowned Artificial Intelligence (AI), Computer Vision and Advanced Analytics technologies. The project aims included:

  • Reducing human inspection and surveillance time
  • Faster defect detection for reduced downtime
  • Improving defect detection accuracy
  • Reducing safety risks associated with driving through operational areas
  • Reducing mundane/repetitive tasks to increase job satisfaction
  • Gain insights into the overall field health through predictive analytics

Industry Context and Challenges

With thousands of assets to manage across thousands of square miles, inspectors had to drive millions of miles and spend millions of hours visiting sites, manually looking for defects and leaks. This work exposed them to significant environmental risks, such as managing methane emissions.

The excessive time required to carry out these inspections is allowing a maintenance backlog to grow. Which means that issues are not dealt with quickly enough, ultimately costing far more to resolve.

Our challenge was to collect data from 240 operational wells spread over 240 acres. This would demonstrate the accuracy and efficiency of the Inspection² software performance at well sites, around electrical equipment and along pipelines, while simultaneously looking for leaks.

Inspection² Machine Learning and Computer Vision technologies

Our customer provided us with an extensive inspection checklist. From this, Inspection² ascertained that they needed a machine learning-based image recognition and visualisation tool that:

  • Processed sensor/video data from UAS flights to determine the condition of well sites, benchmarking against their checklist
  • Presented data through a cloud-based interface, including visualisation of individual assets, overall field health, asset reports and anomaly alerts

To accurately capture static, thermal and methane data and then detect any changes, we used UAV following set flight paths for repeatable and consistent data. From those images, a subset was used to train the AI algorithms.

Multiple parameters were necessary to finetune and measure the algorithm, allowing enhanced accuracy for image decomposition, anomaly size and distance, lighting considerations and so on.

Being a complex project, this process took time. We needed to refine the model through various stages of development. Working closely with the customer’s expert engineering team certainly aided that process, helping to ensure its validity at all stages.

%

accuracy on 35% of the inspection checklist

%

accuracy on 25% of the inspection checklist

AI DNN’s to detect all the well site components

Excellent detection and reporting results

In under four months, using multiple software components including AI, Computer Vision and Geo-Positioning, we developed software that:

  • Supported the ingestion and transcoding of four required types of data – Visual (new format 100-megapixel images), Thermal, Methane and IR
  • 20 AI Deep Neural Networks (DNN’s) to detect all the well site components in the checklist
  • Change Detection solution
  • Customised inspection report template

Despite the time limit of this first operational phase of the pilot project, we achieved 100% accuracy on ten items of the inspection checklist and 80-90% accuracy on seven others. Given more time and data in the next operational phases, we will see the AI accuracy improve even more.

With real-time asset visibility, and by centrally monitoring the status of equipment at remote well sites, Oil and Gas operators have the data needed to eliminate inaccuracies and ensure greater critical asset utilisation, resulting in maximised productivity. Fast.

In conclusion, the project demonstrated that data from UAVs combined with Advanced Analytics have real potential to deliver a commercial solution that is scalable to thousands of assets.

What Inspection² software can do for your organisation

Inspection² develops bespoke software specific to many different industry types. Our software ingests data and uses Computer Vision and AI algorithms to provide automated insights into the condition and usage of your assets. And it all happens faster and more efficiently than ever. Our software can:

  • Review the AI results through a combination of filters
  • Access all annotations, observations and data analysis through highly visual and intuitive inspection reports (web-based, PDF report, 3D viewer) from anywhere at anytime
  • Update work orders via API through easy integration with Asset Management Systems/ERPs such as SAP or IBM’s Maximo

To find out how Inspection² software solutions can help your organisation, please get in touch. We’ll source the ideal solution for your industry sector, saving time and resources, providing an excellent ROI.

After all, it’s natural progression, but with science and technology.

Pin It on Pinterest

Share This