How do you ensure that every step in the production of a high-volume product is really beneficial? Volvo Cars analysed data from the production of 120,000 cars. The result: an entire complicated step was eliminated – while at the same time identifying possible improvements in the production line.
Volvo Cars has invested decades of work and countless resources in building a well-known global brand with values such as quality and safety. Along the way, the company has mastered the mechanics of building cars with the right quality and delivery reliability. The Torslanda plant produces a new car every minute.
From mechanical designers to software developers
"From this perspective, the digitalisation of vehicles has been rapid. In just 10-15 years, we have become software developers in the large, complex system that is a modern car," says Joakim Pernstål, Senior Manufacturing Engineer at Volvo Cars, who is part of a team that equips each fully assembled car with the right software.
At the end of the production line at the factory there has been a station for several years where the car's radar and camera systems are quality assured and calibrated. This is a complicated, expensive and relatively time-consuming process. Joakim Pernstål's team has long been sceptical about the customer benefits. Calibration may sound necessary, but the systems in modern cars calibrate themselves as part of their operation. The question was: could valuable resources be saved by simplifying the process - while maintaining the best possible customer experience?
So the production engineering team, together with the development team and researchers from RISE, embarked on a project to analyse data from the production line at the Torslanda assembly plant.
"Impressive demands on ourselves in terms of customer satisfaction"
Tomas Westund is a project manager in AI and applied IoT at RISE, and the person who led the project from the RISE side:
"It's impressive what level of customer satisfaction Volvo is aiming for. This is not about the safety of the system itself, but about the fact that, for example, the system may need 10 minutes of calibration the first time the customer drives the car before the adaptive cruise control can be used."
By reviewing and analysing data from 120,000 cars produced, the team was able to study the performance of the different calibration stations in Volvo Cars' factories.
"One of the results was that the stations calibrated differently. This means that the car still performs its individual calibration on the road, and it is questionable whether the extra step contributes to the improved customer experience it is intended for", says Joakim Pernstål. This is a new insight. If it was just for calibration, you could skip that step and save time and resources.
In existing factories we can save time and costs
The result can save time and money
"The potential savings for such a station are around SEK 18 million. Although they are already in use in existing factories, when we build new factories - such as the one planned in Slovakia – we can save both investment and space."
"In existing factories we can save time and money. Because we produce such large volumes, this adds up to significant amounts," says Joakim Pernstål.
Tohid Ardeshiri, a researcher at RISE, worked on the data analysis.
"It's easy to get a bit confused when you talk about data analytics, AI and machine learning. But what we have done here is not complicated. We happen to be a team that has experience and understanding of how to process large amounts of data and draw the right conclusions from it. It's not a big deal, it's just that we know how to sift through all the data to find the gold."
For Joakim Pernstål, it's now a matter of finding ways to extract the real value from the conclusions. Even if calibration is not required, the radar and camera system must still be checked. In rare cases, parts that have been fitted do not work or work incorrectly. This needs to be identified and corrected.
"But this is also a good and important insight that we can use to optimise the process. Obviously, we should be putting our energy into improving and checking the assembly process as it is happening, so that errors are eliminated or corrected during the assembly process, not when the car is finished. So now we have development work to do upstream in the production process, which is part of our daily work to improve."
Added value: new external contacts in data analysis
The project also brings added value to Joakim Pernstål's team:
"We already have strong external contacts and networks in mechanical production. This project has also given us valuable contacts in software development and data analysis, which will be essential as cars become more connected and dependent on software. We also have large amounts of data in other areas, such as welding and bolted joints, where such robust analysis can give us new insights into different types of improvements."
How to know if your organisation is ready for improvement through data analysis
- Do you collect data? The main rule of thumb is that if you collect data, you can analyse it.
- What data do you collect? Process-related data such as test data, measurement data, cycle times, uptime and so on, but also data from business systems and maintenance systems... Any type of data can be analysed. Sometimes there is an unexpected correlation between systems that an analysis can reveal.
- Size doesn't matter... You don't need mass production to do data analysis. In theory, you could analyse data from the production of a space shuttle that is made once a year. There may still be parts of the process that are done many times (welds, etc.) that could be analysed.
- ...but it can certainly help. In general, you should have at least 5000 data points to be able to find insights and improvements.
