Changing World of Automation

The drive towards automation throughout the industrial sector began a long time ago, it’s nothing new. The benefits of automation include cost savings and safety improvements, as well as reducing carbon footprint, which is a current focus of many companies.

The goal of automation of the industrial asset lifecycle through software continues today as it has done for the last 20 years. But the most recent game-changer within the ongoing progression is the emergence of Artificial Intelligence software.

In the last five years, AI has matured in terms of its reliability and its affordability so it can be applied to safety-critical missions and industrial assets. AI has opened up completely new ways to deliver automation.  

Inspection² have been developing software to automate the industrial asset lifecycle since 2014 with a focus on Artificial Intelligence software since 2016. The following seven steps deliver an up-to-date method of how you can achieve automation using the latest technology throughout the asset lifecycle.

Seven Steps To Automation

1. Understand the engineering of the asset

The complexity of a bespoke off-shore oil rig is higher compared to that of an electric pylon. For an oil rig, we would look to automate a particular activity e.g. fabric maintenance. For simple structures such as electrical pylons, you can be more aggressive in your automation targets.
Either way, simple or complex, you must start designing and building the automation process of the asset lifecycle with the core engineering principles for the asset in question.
Starting with the technology you have won’t deliver success. Building on asset engineering principles will.

2. Focus on Life Cycle Phase

There are a set of standard phases that any industrial asset will go through during its life. Trying to automate them all in one go might seem logical to deliver a complete packaged solution. However, it’s not where you will get your biggest bang for your buck. Life cycle phases such as Construction or Decommissioning are critical to getting right, but they are one-off and hopefully many decades apart.
Inspections, Maintenance and Audits Checks are where you will see the benefits quickest. Asset Upgrades and Design Changes are more challenging but should have a place in the scope of any automation programme.

3. Data Capture Method

We have dealt with data captured from handheld cameras, helicopters, drones, satellites and mobile phones. All have a place. But if we had to pick one, it would be ‘drone captured’ imagery and data. A drone can fly autonomously. They capture geo-referencing information that makes automation easier, and they can do it faster & more safely than a human being.
The other important consideration is what sensors you want to use to capture asset data. The initial focus will be on visual (RGB) imagery. However, in our experience to automate a process for industrial asset visual imagery is only one part.
To truly automate an asset, you will likely need to incorporate additional sensor data types based on the asset and the scenario. These include; corona sensors, thermal data, methane detection, IR imagery, multispectral imagery, gas detection, etc.
These will require specialised software but are needed to deliver the kinds of full automation our customer requests.

4. What to Capture

What you need to capture is both asset and stage dependant. Whatever the scenario you need a plan.
That capture plan must cater for the different types of asset sites, configurations and setup. A city rooftop cell site will have a different set of capture instructions than a 20-meter mono-pole cell site out in the countryside.
Inspection² ICC (Intelligent Capture Checklist) functionality can cover that. We have a standard configuration tool that lets you set up your capture checklist.
If you are developing this part of the solution yourself, the basic functionality is not complex. There is only one thing to watch out for: this functionality needs to run in remote locations (particularly for distributed assets) where the internet connection is not optimal.
The part that takes it to the ‘next level’ of automation is the bespoke ‘intelligence’ element we have developed within our ICC functionality.

5. Component Analysis

We work on the basis that you need to identify what you are looking at first before deciding what is wrong with it or what needs to be done.
Here is where AI comes into play. Using the latest image analysis AI software technology (called convolution neural networks), the identification of components and objects that make up an industrial asset is relatively straightforward to develop – assuming you are used to developing AI.
There are ways of doing this efficiently and ways of doing it badly. However, the software available today and its level of maturity make the automation of the fifth step very viable.

6. Anomaly Analysis

This step is all about automatically identifying what is wrong with the asset in question. In this case, the software does become more complex since you are replacing a trained engineer

As long as the preceding steps have been followed, this step is feasible. We recommend using a range of different technology and software modules to handle different asset scenarios. These include:

₋ Computer Vision software which is good at image decomposition and measurement.

₋ Geo-Position software to set context and location

₋ AI software

₋ Change Detection software

₋ Risk Management software

₋ And specific asset type tools such as azimuth calculation

It would be naive to trust that artificial intelligence can solve all your problems within this step. Our Head of AI and Computer Vision, Doctor Thomas Moranduzzo, sheds some further light on problematic situations where it is crucial to combine AI with other techniques.

7. Results against Business Rules

The final step brings together the specific data captured on-site with the processed data produced from steps 5 and 6. At this point, everything comes to a conclusion.
We do this through a set of ‘customer-defined business rules’ and ‘engineering output templates’. We take this approach because of its flexibility to configure the application to meet the demands of our different customers.
If you are doing this for yourself and understand the outputs required to answer the asset inspection questions, then ‘old school’ procedural programming will do the job.


There is undoubtedly a lot of software required to deliver industrial asset automation. It took us five years of continuous investment and development to provide that capability via our software. However, given the scale of savings and benefits, continuing the journey to automation is not only possible by using the latest technology, but it is also absolutely worth it.

Adrian Karl

Adrian Karl

Co-Founder & CTO

Pin It on Pinterest

Share This