By Alessandro Chimera TIBCO’s director of digitalisation strategy

The way we use data to monitor and control manufacturing systems has changed. TIBCO’s director of digitalisation strategy, Alessandro Chimera, explains the shape of the anomaly economy guiding operational performance management in manufacturing.

“A stitch in time saves nine” is a well-known and well-worn adage. It neatly encapsulates the reality that early remedial action to deal with a minor fault detected can avoid greater repair costs caused by damage down the track.

Never has this saying rang truer than in today’s high speed, high volume, near real-time manufacturing processes, often described as Industry 4.0, where rapid production can make fault detection difficult.

Without appropriate anomaly detection and rapid intervention, it would be easy for large volumes of a flawed product to be distributed before a problem is discovered, or for expensive manufacturing machinery to be seriously damaged.

Detecting anomalies does not only increase production it can produce considerable savings: machine downtime and possible damage can be avoided along with product wastage and, most importantly, flawed products can be detected before they are delivered to customers.

However, this is not an easy task, even with the availability of massive amounts of real-time data, artificial intelligence (AI) and machine learning (ML).

This data is likely to originate in multiple different formats from different systems and scattered across many different locations.

For example, some valuable data might need to be gathered from products that have already been shipped that have exhibited abnormal behaviour or failed.

To make this data usable for AI and ML techniques, it must first be harmonised and related to the components of the manufacturing process it represents so it can be meaningfully analysed. This requires the use of an autoencoder.

It is not an easy concept to explain in-depth, but simply put, an autoencoder is like the neural network in the brain, which teaches itself to ignore insignificant data.

This is perhaps the best, simple explanation. It takes as its example a digital model of a house. “[A] first autoencoder process will learn to encode easy features like the angles of a roof, while the second analyses the first layer output to encode less obvious features like a doorknob. Then the third encodes an entire door and so on until the final autoencoder encodes the whole image into a code that matches the concept of a ‘house’.”

A good example of the power of anomaly detection can be found in its use by US semiconductor manufacturer Hemlock Semiconductor. A key raw material Hemlock uses for semiconductor manufacture is polycrystalline silicon. It is a particular form of silicon with very low impurity levels, less than one part in a billion.

To be competitive, Hemlock Semiconductor must keep production cost and energy usage to a minimum, consistent with producing reliable products. However, reducing costs means increasing the risk of producing products of unacceptable quality.

By applying anomaly detection to the vast amounts of data generated in its manufacturing process, Hemlock Semiconductor has been able to reduce risk without increasing costs.

The anomaly detection system raises an alert whenever key parameters pass certain predetermined thresholds. This enables remedial action to be taken before product quality is compromised.

The wider use of anomaly detection in manufacturing will not only increase its users’ efficiency and competitiveness, but it will also bring environmental benefits by leading to more reliable and higher quality products and reductions in waste and energy consumption.

For Australia, not-for-profits like the ARM (advanced robotics for manufacturing) Hub, are assisting manufacturers to embrace AI to overcome the present skills shortage. It provides expert advice on AI solutions and scientific and technical expertise.

During its first year of operations, it reported it had worked with over 200 companies in this capacity, seeking ways to maintain and grow their digital capabilities.

Two successful ARM ventures are Verton Pty Ltd and Australian Droid and Robot (ADR) Pty Ltd. Verton took a leap when it reinvented heavy-lifting operations, delivering “safer, faster and smarter crane operations”, but now with the addition of its AI capabilities it is forming international joint ventures and looking for big data opportunities.

This reach into AI was its response to the industry 4.0 call for digital capabilities to aid and improve manufacturing processes, particularly in the integration of advanced robotics and artificial intelligence.

ADR is also demonstrating how AI and machine learning can help Australian manufacturers increase productivity and boost efficiency. It has developed a range of robotic vehicles for mining, industry and for search and rescue operations.

It has incorporated AI and machine learning technology into its products to help Australian manufacturers increase productivity and become more efficient. For example, it has developed a robot that uses computer vision and machine learning algorithms to detect disease in ginger root stock.

Australia’s ARM Hub is really delivering some exciting AI initiatives and looking to the future, while becoming increasingly successful in demonstrating how AI and machine learning technology may assist Australian manufacturers in increasing productivity and efficiencies.

Anomaly detection technology systems are on the rise and manufacturing industries around the world should get ready for their full integration which will help increase efficiencies and production output by significantly reducing errors and product flaws.

Ultimately, this technological progress is leading us to a new tomorrow for us and our planet. As the use of digital twins expands in line with the application of data analytics and ML, every manufacturer will become a data-driven factory, focused on increasingly precise anomaly detection.

When this happens, manufacturers will be able to work more accurately with the added bonus of optimising energy consumption, which will help all as demand and costs for energy increase day by day around the globe.