COPA-DATA Blog

Predictive analytics for food production

Written by Alan Binning | October 2018

Unexpected downtime is one of the biggest threats to revenue in manufacturing. For those in the food and beverage sector, this risk is amplified by the reliance on perishable goods for production. One of the most common causes of downtime is unplanned repairs and unscheduled equipment maintenance, necessitated by the failure or breakdown of machinery.

Downtime isn’t a new problem. Manufacturers have long implemented scheduled maintenance plans to alleviate this challenge. Historically, this would require the implementation of annual or six-monthly schedules to replace and service machinery on the factory floor. But, this isn’t always the most cost-effective option. What’s more, this method has the potential to introduce faults where there were none before, due to improper maintenance work.

 

Health monitoring

Technology has evolved to provide manufacturers with constant insight into the health of their equipment. Rather than inspecting machinery on a scheduled basis, simply because a timetable advises them too, manufacturers are using this technology to repair items that actually need attention — and leaving those that are in good working order to continue operating.

There are various forms of condition monitoring, many of which are enabled by sensors. Sensors allow manufacturers to monitor vibration, pressure and temperature readings, identifying when a machine or part is beginning to show signs of wear. Advances in industrial automation software are also enabling better visualisation of this information, allowing manufacturers to react quicker, and with better insight — even across multiple sites.

By amalgamating operating information from the various islands of data on the factory floor — be that from large machines or smaller components — manufacturers are provided with a factory-wide view of performance. Using this information, manufacturers can identify when a piece of equipment is showing signs of wear and act accordingly.

A predictive maintenance strategy describes repairing equipment based on its actual symptoms, informed by real-time performance data. But, could software predict these failures before the equipment shows any signs of wear?

 

Predicting the future

Recent software developments are enabling manufacturers to identify future faults in machinery, before any signs of failure occur. Using machine learning technologies, data models can inform why and when a machine might fail based upon trends in machine data.

Consider a depositor used in food processing as an example. Depositors are often used in conjunction with belt conveying systems, to automatically fill product cases with an ingredient or mixture. Due to the combined effort of these processes, ensuring the equipment maintains a speed synchronised with the conveyor is essential.

Monitoring software could pinpoint the typical lifespan of the depositor’s motor by examining historical data from other equipment of this type. Combining this information with the typical operating hours of the factory and the production line speed, there is enough information to predict how, why and when the equipment is likely to fail or deteriorate.

In this case, failure to identify a potential breakdown could cause significant downtime. However, even the slightest reduction in efficiency could also skew the accuracy of depositing, causing faulty batches in production.  

This method of condition monitoring is often referred to as predictive analytics. However, an influx of AI developments is making this technique even more effective and advanced, coining the term, prescriptive maintenance.

 

Incorporating AI

COPA-DATA, developer of industrial software zenon, has collaborated with specialists in the fields of data science and artificial intelligence to ensure its software can deliver the highest level of insight for manufacturers.

With partner, Resolto Informatik, COPA-DATA’s software uses algorithms from the field of machine learning that can interpret readings from sensors and actuators to detect correlating patterns in equipment failure. When combined with historical and real-time data streams, it is possible that this wealth of data could grant manufacturers with enough knowledge to completely eliminate unexpected breakdowns.

In a food manufacturing setting, unplanned downtime due to breakdowns can cause staggering financial losses. This is due to the high volume of products being produced continuously, plus the risk of spoilage of unused ingredients.

While traditional, planned maintenance does have its place in food manufacturing, technology has evolved to better inform the maintenance of industrial equipment. Looking to the future, manufacturers have the opportunity to deploy technologies to inform as well as act when maintenance conditions arise. This is thanks to automation platforms such as COPA-DATA’s zenon, which simplifies the process of integration with ERP systems. With companies reporting up to 80 per cent reduction in downtime using these platforms, can you afford not to use of predictive analytics?