Electricity and Control June 2021

COVER ARTICLE

Beyond electricity and control to anomaly detection and machine learning

F orty years ago companies started switching over from pneumatic and relay logic to electronic automation and control systems. At first these systems were all customised or built-to-purpose with little or no scalability and re-use. Over time they have become commercial off-the-shelf systems that are seen and used as commodities in the field of plant automation. Even though a lot of industries still operate manual systems and processes using a combination of manual labour and stand-alone equipment, there is a move to greater automation and systems integration. Manufacturers are looking to the automation industry to deliver additional benefit in the form of business-process automation (as opposed to physical process andmachine automation). This higher level automation is commonly called Manufacturing Operations Management (MOM) or a Manufacturing Execution System (MES), and most automation companies have added components of this to their product offerings. These systems are typically more complex than normal automation systems, as the data exchanged is normally contextualised information shared between different areas or departments of the business. This requires operators to change the way they operate the manufacturing facility. Due to the complexity and risk, these systems have taken longer to become widely accepted and it is only in the past few years that they have become more common. Over the same time, much has been said about the Fourth Industrial Revolution (4IR), the Internet of Things (IoT), Artificial Intelligence (AI) and Machine Learning (ML). These concepts, even though expressed in simple terms, sound utopian or futuristic to most manufacturing companies. Most believe they are not ready for any of this as they operate manual systems or have only islands of

automation but no integrated data acquisition system. What many companies fail to recognise is that without this real-time data visibility, their timeous decision-making ability is lagging and so is their market competitiveness. Making a decision to change hours after the event as opposed to seconds after the event wastes time and resources and makes companies less competitive. In the long run, this lag in decision-making can lead to ruin. Therefore, having actionable information in real time is critical to improving the efficiency and effectiveness of operations and ensuring long-term sustainability of manufacturing. Real-time information, once available can also be used for ML. By applying the appropriate algorithms to a historical dataset, companies can identify anomalies in equipment and process performance in real time. This enables them to predict upcoming quality issues, impending equipment failures, imminent process deviations and others, so they can be prevented. The difference between the automation systems of forty years ago and ML systems today, is that the time for them to evolve to become commodities has already passed and the time for them to be widely adopted has decreased considerably. There are companies today (small, medium and large) that have already invested in and are using ML solutions in their plants. Thinking that you have time before these systems prove themselves and that they present no threat to your business is a fallacy. Iritron can advise companies on MES, MOM and ML systems. □

For more information contact Iritron. Tel: +27 (0)12 349 2919 Email: gerhard.greeff@iritron.co.za Visit: www.iritron.co.za

Actionable information in real time is critical to improving operational efficiency and long-term sustainability of manufacturing.

Electricity + Control JUNE 2021

3