Everything you wanted to know about Machine Learning and Artificial Inteligence (but were too afraid to ask!) By Graeme Wright

The saviours of energy, oil and gas companies have been long been heralded as data analysis, and with the development of machine learning, anomaly detection, and Artificial Intelligence (AI) it is said to hold the key to increased resilience, better asset management, improved customer experience and service. It almost seems too good to be true. But beneath the hype, few companies actually know which ones to apply to which problems or how to use the resulting insights to move beyond predictive analytics to deliver optimised value.

These buzzwords have become increasingly popular for a number of reasons. Importantly, because the solutions for insightful analytics are now cheaper and more accessible than ever, something which does not seem to be slowing down. The computing power can be available over the internet with no capital investment, making this possible and available for any business. Improvements in both algorithms and memory capacity have also contributed to the popularity of these technologies. Huge quantities of data can now be accessed and dealt with in a more efficient and effective way.

However, companies are still unable to distinguish between the technologies and identify which is the best to use. This often stems from a lack of understanding about the original issue. I often find this comes in one of two ways: a wealth of information, but uncertainty on how to use it, or a lack of data and a rush to find a solution. The first can lead to requests such as ‘if I give you an extract – tell me something I don’t know’, and the second, ‘how much data do I need?’. Both of these indicate that they do not fully understand what their business issue is. Once this is achieved, they can look at what solutions are available to help them fit their problem and achieve their goals. There are subtle differences between the technologies however, which I will explore below.

Artificial Intelligence (AI)
AI has the potential to streamline processes within organisations. Its primary use is in automating administrative tasks, call centres, and aspects of Finance and HR. Chatbots, for example, are already becoming used more widely. Advances in AI will allow companies to accumulate their knowledge and skills onto a platform, which can then be used by a Chatbot to transfer that information to an engineer onsite. Realtime aid and advice from Chatbots will allow engineers to work in a more efficient and productive manner, as they no longer have to rely on other engineers. AI can be used to identify anomalies, and what the next best course of action is to deal with these issues. This helps the engineer when onsite, but also provides preventative measures that will help to reduce issues and problems in the long run. In advanced cases, AI is now being applied to different imaging solutions such as CCTV, ultrasound or even ground penetrating radar, acting as highly skilled analysts to identify faults or dangerous situations – without getting tired or missing key signals.

However, training AI can be an issue. It is only able to detect problems on information that it has been trained with, and so is taught what is normal and abnormal. Time taken to train AI can also pose a problem, especially when the right information cannot always be accessed.

Machine learning
While AI can only be trained with specific information, machine learning is where algorithms are used so that a machine learns characteristics for itself from a vast wealth of data, without human involvement. This means that the machine is able to identify and recognise what data patterns lead to a known problem. Importantly, as the machine is exposed to the data, it is able to adapt. Through the use of machine learning, organisations are able to protect themselves from unknown and complex issues, as well as being able to identify potential opportunities, with the algorithms solutions that machine learning provides.

Predictive maintenance
Predictive maintenance uses AI and/or machine learning. It focuses on predicting when and why a machine or part of a machine might fail or need maintenance work. This is obviously useful for companies as it will make them situationally aware of their operations and put measures in place to combat any issues. Economic and time costs will therefore be reduced and aid the efficiency of the company. From existing Industrial Control Systems (ICS) there is already a vast amount of data that can be accessed by organisations. Whilst this data is diminished in value because it is kept in organisational silos and often much of it is only available on site rather than at the centre, it is still beneficial to companies. Coupled with the reduction of communications costs, particularly with sensors and wireless solutions, I believe this will cause more companies to begin to deploy predictive maintenance.

Moving to an optimised world
With all of the insights comes the next challenge – optimisation. Too often, I have heard the complaint that staff on site have a list of jobs to do that they just can’t complete; so, they don’t – they do what they can! However, what is best value, how do you solve the complex combinatorial problems of task scheduling based on the best value to the business? Optimisation can quickly become unsolvable with traditional computing approaches within an acceptable time and cost. Yet, breakthroughs inspired by Quantum computing has started to make massive advances in cost and time taken to optimise the insights to drive value to the business.

Final challenges
Skills shortages and changing practices in the workplace, however, are still limiting factors in using these technologies to benefit organisations.

When technologies are used effectively, they can be best used to drive business value. The data that energy, oil and gas companies have access to has great value, and so companies must adapt and use this readily available resource. By not doing so, they could damage or stilt their business’ future and opportunities.

Fujitsu UK & Ireland
Graeme Wright is Fujitsu’s CTO for Manufacturing and Utilities. Fujitsu UK & Ireland employs over 9000 people and promotes a Human Centric Intelligent Society, in which innovation is driven by the integration of people, information and infrastructure. Fujitsu enables its customers to digitally transform with connected technology services, focused on Artificial Intelligence, the Internet of Things, and Cloud – all underpinned by Security. Its customers cover both the public and private sectors, including retail, financial services, transport, manufacturing, government and defence.

For further information please visit: http://uk.fujitsu.com