Machine fine-tuning, our first project in manufacturing

For the last few years everyone talks about the importance of advanced analytics for manufacturing as a next step after lean and Six Sigma programs and what great potential it can unleash. So when we were in front of our first project we were naturally very excited and curious what can be done. The outcome of the project exceeded our expectations both in terms of data modelling and more importantly business results for our client.

A leader among manufacturers of plastic components for the automotive industry wanted to fine-tune one of their injection moulding machines. After several months of kaizen they had reached a ca. 1% scrap rate, but were unable to improve it any further. The ultimate goal being a zero scrap rate – a holy grail of manufacturing. They decided to reach out to an external partner experienced in advanced analytics, which is how we got into the picture.

Soon we learnt that injection moulding technology, which has over fifty parameters to be set and thousands of process sensors recording data, is one the most complex tasks in manufacturing. As with any analysis, the first step was to link the data with the machine’s settings used for each production cycle to a database. Then we added process sensor data as another layer.

Blind example of few data fields for 1500 cycles

Since the goal was the zero scrap rate we obviously had to add the data from the system capturing the quality control outputs. This turned out to be slightly more complicated than we expected because the data came from a completely independent system. However, using the right technologies and applying some tricks we have learnt so far, we were quickly able to link everything together.

Once we had all the data ready in a nice format we started with the modelling. As always, this was the most interesting part of the project. The complexity of the moulding process as represented by the input settings and sensor data was enormous. On the other hand, compared to client-related data we usually work with, the data volume was huge and the data were representing exactly what they were supposed to. This enabled us to use state-of-the-world modelling techniques, which are extremely powerful in detecting the famous signal from the noise. Together with the client we run few experiments to create variability in the data and finally let the algorithm recommend the optimal settings.

The algorithm suggested optimal settings and process window for the machine that eventually led to a zero scrap rate. This had two important benefits for the plant manager – reduction of scrap costs and most importantly no need for visual inspection of the output quality. It is extremely beneficial for a business influenced by a minimal unemployment rate when they can fully automate a process and use the human labor elsewhere. In our project that actually meant annual savings of ~100 ths. EUR per machine.

Great impact, fully delivered on its promise!

Kristijan Fiket, CEO of Aures, former Vice President of Operations of Automotive Lighting

Having this experience with advanced analytics in manufacturing convinced us that not only the potential everyone talks about is real, but it is also a wonderful analytical playground to be in. We hope that more projects like this one will follow soon. So stay tuned!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.