This article discusses how manufacturers can benefit from the latest machine control technology, such as the development of machine learning models capable of predicting that a critical element of manufacturing equipment is at risk of breaking down and ordering spare parts in advance.
What if you had access to a machine learning model capable of predicting the breakdown of a critical element in your manufacturing equipment – and it could order spare parts in advance? Or what if it were possible for an industrial controller with built-in artificial intelligence (AI) to use pattern recognition to predict an unwarranted deviation on your filling line that would otherwise have gone undetected?
Although these examples sound out-of-reach, they exist in the present day as applications for AI on the factory floor. AI as a research discipline is several decades old, but thanks to technological advancements and an increasing focus on data, it is also becoming one of the most promising mechanisms of Industry 4.0.
Discussions around AI in manufacturing are gaining traction. It is already used in many industries, aiding companies in the analysis of large amounts of data to improve their R&D and plant management processes. However, AI can also offer significant advantages on the machine level, allowing manufacturers to re-think their processes to be more flexible and adapt to any potential change.
Transforming production and implementing AI can help manufacturers to prolong equipment longevity and detect unforeseen events to prevent failures. This reduces the risk of equipment damage and downtime. It shows a manufacturer what needs to be done to improve their system, providing practical requirements to noticeably boost overall equipment efficiency (OEE). An enhancement of just a few percentage points can result in significant efficiency gains and cost reductions.
Highlighting the significance of operational excellence
Operational excellence is required to maximise capital expenditures. But at the same time, manufacturers face challenges when moving away from traditional high volume, low mix production runs, to more complex high mix, low volume production – with demand linked directly with production.
In addition, legislation and retailer mandates, for example in food and beverage, is putting greater demands on manufacturers to ensure that labels printed are accurate and contain all the data required, including readable barcodes. Late-stage customisation includes, for instance, adding variable data to labels or packaging that are pre-printed. This allows the manufacturer to customise product labelling for certain markets, customers, or products without having to stock individual packaging materials for each possible variant. In addition, the flexible production of smaller batch sizes places new requirements for automatic and fast changeovers.
Manufacturers can achieve operational excellence by implementing ‘AI at the edge’. However, they also require flexible and autonomous production support, as well as IoT automation solutions from data collection to intelligent production for a seamless integration of IT and OT worlds. In the case of the advancements required for Industry 4.0 – such as predictive maintenance and networking – the use of adaptive algorithms offers enormous potential. Many companies are realising that AI presents an opportunity to increase not only the OEE – and therefore combine reduced costs with increased productivity – but also to improve the analysis of data to support continuous improvement programmes such as reducing waste or process operations variability.
According to a study by Aberdeen Group, although OEE values of 89% have been achieved by leaders of the industry, many of the traditional systems currently in live usage have been generating figures of around 74%. But what if we look beyond this, and add in AI solutions for automation? If the quality can be improved and predictive maintenance can be used to prevent machine downtime, much more significant efficiency gains can be made. In any case, what really matters about getting OEE information throughout the entire process is what a manufacturer does with it – and it can be used to tackle the identified pain points.
Manufacturing with the edge and cloud
What do we mean by AI at the edge? You can describe small-data machine learning as ‘spinal reflex AI’. At this level, lines and devices are monitored with real-time sensors, and data is collected and processed at high speed for fast detection of abnormalities. To translate information into action, manufacturers need efficient control and monitoring for a more natural, proactive relationship between operator and machine, deploying technology that brings together all areas of automation, including logic, motion, vision, safety, and visualisation. This helps companies increase productivity in a very flexible way.
Processing big data in the cloud can be described as ‘cerebral AI’. This requires open and secure standards such as the MQTT protocol and the OPC UA communications standard for safe and easy transformation of machine and system data into high value information.
Many of the AI solutions advertised on the market, which are often cloud-based, have significant requirements in terms of infrastructure and IT. These solutions also work with an overwhelming amount of data that is laborious and time-consuming to prepare and process. The question of added value often remains somewhat murky for providers, who cannot determine whether or how investing in AI will provide a return. The fact that system designs for the production industry are generally both complex and unique is another contributing factor.
Given these conditions, how do we go about designing and integrating AI that creates tangible added value in the production process? While the cloud is best suited to deal with big data and manage massive long-term analytics, AI at the edge is crucial for real-time applications. This approach offers more flexibility and faster response times, so production environments can get better use of data analysis at the edge rather than depending on cloud computing. Instead of laboriously searching a huge volume of data for patterns, in addition to the processes that are running, it’s important to tackle things from the other direction. Technology is needed where the required AI algorithms are integrated into the machine control system, thus creating the framework for real-time optimisation that is truly on the edge – at the machine, for the machine.
What does AI at the edge provide?
With AI at the edge, manufacturers can better control complexity and security. Although the scope of data remains relatively large, organisations require fewer resources in terms of hardware, communication infrastructure, and processing capabilities at enterprise levels.
Technology that offers seamless integration to the cloud directly from the machine controller with embedded, secured IoT protocols is needed. With AI at the edge, organisations can expand their processing and analytics capabilities with AI algorithms at machine level, benefitting from faster and more efficient information for their decision-making process. Furthermore, users can analyse and act on the process in real-time, exactly where the action happens, improving performance autonomously with AI and machine learning.
By harnessing AI at the edge, companies get the most out of their cloud computing with preprocessed and aggregated data at the machine level, reducing the IT infrastructure required for an optimum data flow.
How does AI at the edge work in action?
In a packaging machine, for example, a machine controller offers synchronised control of all machine devices and advanced functionality such as motion, robotics, and database connectivity. A machine controller equipped with AI fuses machine control functions with AI processing in real time. An AI controller features adaptive intelligence, which brings it closer to the action and helps it learn to distinguish normal patterns from abnormal ones for the individual machine.
This type of solution is primarily used in the packaging and production process at the points where the customer is experiencing the greatest efficiency problems (‘bottlenecks’). The processes gain intelligence based on previous findings and improvements that have been made, subsequently driving holistic optimisation of the entire manufacturing process.
A good example is a bottling application: the bottles are carried via a conveyor belt and filled. The AI controller learns what a normal situation looks like when no failures occur. In case of a disturbance, such as blockage causing selective friction, anomalies will be detected by the AI controller. The filling process will shortly pause and resume when the process is stabilised. This prevents any failure from taking place – for example, spilling liquid on the packaging line – which in turn helps reduce machine downtime and waste.
AI is likely to revolutionise science and processes across industries in the coming years. However, it can also be implemented by manufacturers to transform their production processes on the machine level through machine controllers equipped with AI – and this can happen right here, right now.
Predictive maintenance can be achieved with artificial intelligence at the edge, which also allows manufacturers to better control complexity and security. Even small improvements in OEE can result in significant efficiency gains and cost reductions and drive real value for manufacturers in the years to come.
— Lucian Dold is general manager Product & Solution Marketing at Omron EMEAThis article was first published on EE Times Europe