Digital technologies are providing new training solutions for the automotive industry. Nick Holt reports

Professional performance training with Tobii Pro G

Eye tracking systems are now being successfully used to create training programs that reduce errors in manufacturing operations. Mike Bartel, senior research developer at Tobii Pro Insights provided an overview of what this technology offers: “Eye tracking has been around for over 30 years but it’s only in the last five years that the technology has become small enough and unobtrusive enough for it to be used the working environment. With our latest generation wearable eye tracker we are able to set up a pair of special glasses on an employee, calibrate them and have them complete their professional tasks; all the time we are measuring where they are looking, what they focusing on, what might be distracting them. We can identify where the safety hazards are they haven’t noticed, any inefficiencies in their process. We can then use this data to form best practices to implement into training protocols. So we take what we have learned from directly measuring employee attention and apply it to improving the employee experience, improving efficiency and making a better overall job performance.”

The system features an HD scene camera is mounted within the frame of the glasses, pointed outwards and capturing a widescreen view of the world. Inside the glass’s frame are high-speed sensors and illuminators that take 100 pictures of the wearer’s eye every second and then plot that data onto the video that is being captured by the scene camera. “It allows you to see the world through the eye’s of the employee,” said Bartel.

Keeping an eye on qualityThe company is working with automotive manufacturers on a few different components of the process. These cover the painting process, quality assurance and the assembly line process. Bartel explained: “Using Pro Glasses 2 we are able to measure where people are looking as they are visually inspecting the paint finish on a vehicle while it’s on the line. From this we are able to couple of things. Firstly we are able to create a set of best practices for the pattern of eye movement that is associated with a good performance on this particular job. Then we’re able to train people to model that behavior as their learning process. [sam_ad id=17 codes='true']

The company claimed automotive manufacturing clients in Japan had reported decreases in visual inspection errors after implementing eye tracking into their processes. Bartel noted that this covered creating both the set of best practices and implementing them directly into how workers were trained on the line. Another important feature of these automotive studies was to cut down the time it takes to train a worker. In one particular study their goal was to cut back training time for a specific function from six to four weeks.

Bartel offered more detail on how the eye tracking data could be adapted into a behaviour model: “We like to have an objective performance metric on quality, so we are able to do a secondary inspection of the car an determine how many errors, things missed, there were in the initial inspection. We can then go back to the eye tracking data and see the visual pattern of behaviour that is associated with this substandard performance. On the other side of this if we have the eye tracking data from someone who has done a very good inspection (no errors) in a reasonable time frame we can look at their visual strategy. The revue process involves looking back at the eye tracking video, which shows everywhere they look in real-time, plus quantifying that using an attention matrix; what is the optimal length of time to spend viewing elements of the car.”

Staying focused on the jobAnother example offered was with US aluminium casting company H&H, Bartel noted that the eye tracking data revealed that when spillage occurred the eye movement’s of the workers involved were unfocussed and sporadic, they hadn’t taken the time to still eye, head and body movements in order to focus fully on the task. “We took this data and implemented it into the company’s training procedures and took the video of the pours involving spillage to show new employees what it looks like when a mistake is made,” he added.

“It’s not as simple as just showing a trainee a video then expecting them to be experts, there’s much more to it. There are other factors involved such as cognitive processing and motor behaviour, also characteristics that people develop over time as they are doing a particular job, so the visual behaviour is just one piece. However, it is a very important piece. In addition to capturing the eye tracking data we will also play back the video to the expert subject and have them narrate their experience, talking through everything they are doing, offering explanations of why they doing something in a certain way and how to avoid errors. This is our way of layering the visual attention data with the cognitive processing piece,” explained Bartel.

The virtual solution

AMS spoke to Stefan Stuering of Livingsolids discuss how virtual reality programs are supporting training on automotive assembly lines.

Could you provide a brief overview of the system you offer?

Our system allows you to import 3D CAD data, so you have to ability to import the whole car into our software. We have links to the planning systems of vehicle makers, such as VW and Ford, so we can load the process planning information where they refine the work steps to assemble the vehicle at each station along the line. With this content we build a training tool in which you can see the whole assembly process, step-by-step and the structure of the line with the ‘borders of the work stations, the teams, the supervisor areas, etc. You can then navigate to a specific work station on the line and view the car in the state of build at that point.

There are then four levels of increasing difficulty where you learn the [assembly] steps. In the first level you click on the highlighted part then you can see how this is assembled, then in level two you have to remember where this part has to be fitted. You see an icon of the part and the corresponding tool, next you have to click where the part should be fitted. In level three you have to select the correct part and tool from a list and remember where it should be fitted and at level four there is no information, no text instructions. The aim is to provide the knowledge needed to perform the task and help the worker to remember the stages, so they are prepared before they even get to the real vehicle.

Is this intended to deliver all the training need before they go on to the line, or is there still a need for actual ‘physical’ training?

It depends on the customer and the type of workplace they operate. Some will decide on completely virtual training, others where the operations are more complex, requiring additional skills will also do training on the real equipment. Often the equipment and practice assemblies are located in a separate training area.

Is this type of virtual training tool used in any other areas beyond the final [automotive] assembly line?

There are some assembly type operations in the bodyshop, for example mounting and adjusting doors, hood, trunk lid and wings, and this is something we also cover. On the main assembly line we also cover drivetrain and cockpit assembly operations and there is particularly high demand wherever wiring or electrical connections are part of the process, as this can be a complex operation for workers to remember with a lot of connections, clips, etc. required to be secured. We often do the entire process from body-in-white to the delivery of the [finished] car.

We’re also covering the assembly of electric motors for the new generation of electric vehicles; this includes the internal parts such as stators and rotors.

How do you see this type of virtual training technology developing in the future?

We are evaluating a feedback glove for a potential physical interaction with the virtual program. What we are continuing to develop is the authoring process since the time span between the [production] data being finalised and the need to begin the training is very short. So, we have to be very quick in setting up the training program and keeping it up-to-date, this is because in the launch phase the client is often carrying out a lot of work reassignments as they try to optimise the [production] line.