Quality and AI: prediction in production

While a processing step is still in progress, a new artificial intelligence (AI) is supposed to predict with a high degree of certainty whether the component will ultimately fulfil the quality specifications. This is the promise of AI solutions that are being developed at the Fraunhofer Institute for Machine Tools and Forming Technology IWU in Chemnitz. According to the engineers, they represent an improvement on previous in-line testing systems. Where AI is used, time-consuming ejection for testing purposes is no longer necessary. Instead, it is about quality forecasts - and these can be integrated into many industrial production processes, usually even in conjunction with existing, low-cost sensor technology.

However, AI can also be used for optimisation purposes. For example, process input parameters can be controlled to avoid rejects from the outset. Or to reduce energy consumption in production without compromising quality.

Where quality prediction with AI is useful

Typical future areas of application for the newly developed AI solution are processing steps such as

• Drilling,
• Turning and
• Milling
.

When drilling, for example, the speed, feed rate and measurements by a vibration sensor provide information about the expected qualitative result. This makes it possible to obtain a quality statement for deep-hole drilling, for example, without having to destroy the workpiece for a measurement.

KI for the quality of hot forming

In the field of metalworking, the use of AI has proven itself, not least in hot forming. In this process, the workpiece is heated above the austenitising temperature - i.e. around 880 °C - in the furnace before pressing. As soon as the desired target temperature for an optimum hardening result is reached, the hot sheet is placed in the press by a handling system and formed. A martensitic structure is created and the material is thus hardened.

As product quality is the main focus of this energy-intensive process, the furnace temperature is often set higher than it needs to be. By predicting the foreseeable oven temperature, the AI provides data-based assistance that can be used to regulate the oven temperature.

Injection moulding: how AI reduces rejects

In injection moulding processes, the AI monitors special parameters such as the temperature of the mould, the rotational speed of the conveying screw for the granulate, the melting temperature, the holding time of the mould and the cooling time. Timely countermeasures in the event of an unfavourable quality forecast therefore help to significantly reduce rejects.

100% inspections, low number of training data sets

In all application scenarios, the AI can be used directly in the production process (inline) to monitor the entire batch. It therefore enables 100% inspections. Random testing alone is now a thing of the past.

For the training of various AI models, many applications require a two-digit number of data sets, supplemented by expert knowledge about the process. During operation, the computing power of (local) edge computing is often sufficient.

Scientific contact:
Dr Marco Schumann
Department Head Data Mining and Artificial Intelligence
Fraunhofer Institute for Machine Tools and Forming Technology IWU
Reichenhainer Str. 88
09126 Chemnitz
Tel. +49-(0)371-5397-1652
Email: marco.schumann@iwu.fraunhofer.de
URL:  www.iwu.fraunhofer.de

Article from "medizin & technik" from 03/07/2023