Contact us

AI revealed a defective product in time: “damage worth millions was avoided”

04.04.2018| Article

This article was originally published by Tero Lehto on a technology magazine 'Tekniikka & Talous' on March 21st 2018. You can read the original article in Finnish here.

Artificial intelligence has only started making its way to factory production, but machine learning driven analytics is already producing interesting results in pilot projects, and the first factories have adapted it in production.

Emil Ackerman, CEO of Quva, a company specialized in developing software especially for processing industry enterprises, tells that they have built both pilot and production stage realizations for quality assurance, anticipatory maintenance and for increasing the efficiency of production.

Experts and researchers talk about machine learning algorithms, but many also call them artificial intelligence. What is more important than the term, is that analytics can develop with the help of data that has been collected earlier.

Ackerman tells that the most impressive feedback they have gotten this far concerned a manufacturing defect that was detected before the product could get to a customer.

“The feedback stated that the company avoided damage worth millions of euros because the defective product did not get to an end user.”

Quva predicts quality-related issues on the basis of process or raw material data instead of collecting samples.

The company’s public customer references include a list of enterprises in the automation, metal and processing industries, such as ABB, Outokumpu, Outotec, SSAB, UPM and Valmet as well as in the energy industry, the main grid companies Fingrid and Vaasan Sähkö.

It is about bringing the raw material to the production line and processing it. At the end of the production line, we get a product that will be delivered to a customer or to further processing.

The purpose of analytics is to reveal defects in this chain.

A defect in quality may involve a badly printed image on the product package, an image that does not stick to the package well or may fall off, or in the worst case, the package could release substances that it should not release. The product can be ruined or cause danger to the user.

“Quality is an important part of the product image for food or fragrance products, for example.”

Nowadays, consumers often complain in the social media and as a result, the defect of one product can easily get lots of publicity, which can cause significant losses to a company.

Ackerman explains that machine learning algorithms have been used for determining control limits that have enabled the best production result instead of each operator using the machinery with their own settings.

The more data is available, the better the predictions and other analytical results are.

CONTACT US


Quva Oy

Business ID: 2348506-3

Address: Sumeliuksenkatu 18 B
33100 Tampere, Finland

Electronic invoice address: 003723485063

Sales

Emil Ackerman
+358 45 2086 816
emil.ackerman(at)quva.fi

Other contacts:
firstname.lastname(at)quva.fi

Order demo and other inquiries

äLä KIRJOITA TäHäN MITääN

Subscribe to newsletter