Fraunhofer IAO
Fraunhofer IAO

Brain-Computer-Interface for Production Work

In some cases a complete automation of the quality control processes is not possible. Manual visual inspections in particular are demanding cognitive tasks for the human. Such tasks require extreme concentration over a longer period of time. In such cases, a Brain-Computer Interface (BCI) can proactively support the human during the inspection process.

Challenge

Manual visual inspection is a common method of quality assurance in industrial processes. This is usually a visual inspection of a product for any kind of errors - such as scratches, dirt deposits or assembly errors. Since, visual inspections are mainly carried out by humans, they are prone to poor efficiency. This can be caused by fluctuations of concentration, fatigue or distraction. Furthermore, monotonous visual inspection processes have a negative effect on the health and subjective well-being of the workers. Particularly in the case of quality controls that cannot be fully automated, it is important to clarify how the human can be supported.

Methodology

When people detect a mistake or perceive something surprising, this leads to an attention response that can be observed to varying degrees in the person's brain signals, depending on the level of cognitive workload. Even if the reaction to errors can differ between different tasks (e.g. motor or rather abstract tasks), it is still universally recognizable. Electroencephalography (EEG) is an excellent method for measuring this specific reaction in the brain in a time-sensitive manner. Even faster than if the user is able to perform a physical reaction. Using machine learning methods, such specific brain responses to errors are captured by detecting the attention level in real time. We take advantage of these methods to proactively support visual quality inspection tasks.

Results

In a feasibility study, we developed an EEG-based BCI that automatically detects the recognized errors of the human during a visual inspection task and rejects products that are prone to defects with a mean accuracy rate of 85%. The BCI-based error detection can be used to carry out or initiate certain corrections during visual inspection or control tasks. The worker can be proactively supported during the manual visual inspection task by measuring fluctuations in the attentional level to prevent work-related errors. For example, the BCI-system can indicate a reduced attentional level and suggest a pause. The worker can thus be actively relieved by knowing the level of attention in order to make work more human-centred.