Tetra Pak goes digital for quality issue resolution

Using machine learning and domain knowledge can give insights to prevent critical failures, predict machine errors and accelerate response times.

Machine learning involves large amounts of historical data about what happened before a failure so the machine learns and predicts a problem up to a week before it happens.

Condition monitoring uses a limited amount of real-time data and failures can be predicted a month in advance.

The firm said these services would be available where they make sense - so not unique to Tetra Pak machines but based on data generated on the equipment.

Predictive maintenance​

Alex Bromage, food safety and quality business manager at Tetra Pak, said it thinks about digitalisation in several areas.

“We think about it as a connected workforce, providing connectivity to our people out in the field,” ​he told FoodQualityNews at the Global Food Safety Conference in Houston, Texas.​

“Advanced analytics and connected solutions, where we are talking more like the Internet of Things, actually provide a stream of data between parties and getting more connectivity to then realise the big data sets that we do the advanced analytics on.”​

Bromage said it uses advanced analytics, connected to customer machines, to do predictive maintenance or machine learning.

“Predictive maintenance helps us identify when a part will fail before it actually does fail, change it and then we can drive it in a different way than what we have done previously,” ​he said.

“Machine learning is where we take static databases, not live feeds of data, we’ll feed those in and obviously it is a bit more machine-intensive then other things, to test certain hypothesis to try and find the relationship between the different datasets that we have and the performance of a component in the field.​

“What the parameters are that we need to know and can we learn better how to predict the failure? So we have an online prediction of failure and an offline prediction of failure and those are the two main routes we take from a data analytics point of view.”​

All equipment at a customer plant can be connected to the Microsoft Azure cloud system managed by Tetra Pak.

The firm started a six-month pilot in January 2016 for 17 lines across 10 customers in Europe and the Americas.

It said downtime decreased by up to 48 hours for each line saving up to €30,000 for the customer.

Microsoft partnership​

Tetra Pak is also using augmented reality through a headset manufactured by Microsoft called the HoloLens.​

The ambition this year is to deploy 200 headsets with most going to the Americas.

Bromage said it has an order for 50 headsets that it will be sending in the coming months.

“What that enables us to do is equip our field service engineer at the customer site with the HoloLens and whereas before we would fly a specialist in if there was an issue the local engineer was unable to deal with, now he can connect to the specialist who is still in the office via Skype using the HoloLens,” ​he said.

“The specialist is able to see everything that the field service engineer can see and the specialist can mark or augment the environment that the local engineer is able to see to show him documents, instructions and highlight parts of the machine with a pen and they can have a simple conversation using Skype.”​

It will be used in most contexts going forward, said Bromage.

“We will deploy it in conjunction, to begin with, with our existing service contracts to customers. So where customers have an agreement for on-site engineers or a certain service level from us the HoloLens will then be on site and available for use to that Tetra Pak engineer that is assigned to that customer,” ​he said. ​

“When we are deploying new machines, particularly new systems for new markets, we will make sure it goes with that. So we don’t necessarily have to spend a lot of time moving specialists around all over the world.”​

Industry reception​

Bromage said the difference between augmented and virtual reality is with virtual reality you are cut off from visual space around you and watching a screen.

“We are looking to use it in training going forward but with augmented reality there are already industrial applications, the business case has been proven and we can only see that expanding,” ​he said.

“I think augmented is still quite new, it has emerged and it has already been taken up by industry quite well, particularly in this sort of application. Virtual reality has had a couple of cycles already now it is biting a lot better but there is still a problem in terms of your trading off mobility versus performance of the headset. ​

“That is why we are focussing first on getting the training systems built into the iPad into the iOS environment and then we are hopefully going to work in the virtual reality based world.” ​

Feedback from trials and pilots with customers have been positive, said Bromage.

“We have it deployed already in a couple of scenarios and the customers see it is of value. It is saving us all money and time which is good but it is also allowing us to increase the performance of our service level because we can respond to issues faster than having to wait for someone to physically arrive on site,” ​he said.​

“The food industry is traditionally a conservative industry when it comes to technology and that is clearly something that is changing. We see that from the conversations here versus last year. Could it be faster at adopting new technologies, of course, but there are risks associated to that so I understand where it is right now and why it is how it is.”​