The automotive industry is changing, how are manufacturers keeping up?

How will manufacturing keep up with the transition?

When explaining Elon Musk’s vision for manufacturing at Tesla, Tom Mueller the Chief Propulsion Technology Officer at SpaceX said “the major cost of the car is not the material in the car, it’s the factory that builds the car”.

For the product to move forward the machine that builds the product needs to keep up with the technological improvement. There is an opportunity to improve manufacturing considerably by making a greater use of technology. Increasing automation, using machine learning and harnessing the data can increase the efficiency, quality and flexibility required.

If we can build a car that can navigate a busy city street and deal with the unpredictable behaviour of fellow humans. Then we must wonder is it not also possible to build robots that follow predefined tasks in a controlled environment?

Of course, that’s certainly oversimplifying the problem at hand — human dexterity is remarkable. Machines find it really hard to pick up odd shaped objects or attach the clips on wiring harness -but there are still a lot of processes that can be automated or machines used to augment human dexterity and intuition.

Machine Vision

Chihuahua or a muffin? Easy for us to see but for machines, it’s a challenge

With the help of neural networks, machine vision has developed considerably over the past decade. With so much data at their disposal, todays machine vision systems can classify and detect objects at a high level of accuracy. Machines are now able to distinguish between some very closely resembling images. Take the above picture for an example of chihuahuas and muffins with strikingly resemblance. Until recently Artificial Intelligence (AI) struggled to differentiate accurately between images with this kind of similarities. Reassuringly most of today’s popular computer vision tools available online were able to correctly identify the food from the pet, though Microsoft’s system struggled with this particular muffin.

Looks like Microsoft (msft) still has some learning to do.

Quality control and inline part inspection can be considerably improved by using tools such as Google’s Cloud Vision platform. Camera systems at different points in the production line monitored by computer vision can spot quality issues and send parts to be reworked before they get to the end of the line. A lot of automotive OEM’s have been investing heavily in developing computer vision technology for their autonomous vehicle development, that same technology focus should be used in manufacturing.

Automation

Automation on an assembly line also has a lot of potential. Most of today’s manufacturing is semi-automated, a large portion of vehicle assembly is performed with the assistance of industrial robots. Human operators work in conjunction with heavy machinery to assemble the vehicles, humans providing the flexibility and agility while the robots provide endurance and consistent repeatability. Today’s cars may drive at record breaking speed but the manufacturing lines used to produce them are still at a walking pace.

Designing products and manufacturing for automation can increase the amount of work that can be done by machines. Tesla is one of the new manufacturers that has placed its bets on a fully automated production line with the Model 3’s production line likely being the most automated production line of any modern vehicle. Greg Reichow, Tesla’s VP of Production until July 2016 wrote in his Wired article about the company’s manufacturing that “Some of the robots moved at such high speeds that their arms needed to be built from carbon fibre instead of steel.”. The current speed at the Model S and X line is about 5 centimetre per second, by increasing automation Tesla expects to increase the speed 20-fold to about 1 metre per second, that’s a slow to a medium walk speed. This edge in automation will push the throughput and efficiency of Tesla’s manufacturing ability, allowing it to leapfrog established giants.

Tesla Fremont Factory

Data

Data as they say is the new gold, and we’re currently in a gold rush to collect more of it. It’s useful because with enough data about a particular area, unconventional relations and trends can be observed. This insight can give an edge to the company. Companies like Amazon collect a huge amount of data which they use to offer a better service to their customers. They can predict which products are likely to be bought in a certain area and stock the delivery van so when it’s actually bought, it’s already on its way. Netflix captures data about every scene viewers watch and which episodes are repeated or series binged on. This gives them an unparalleled insight into viewer desires and habits which they use to create their own original content.

There’s a ton of data that’s collected at a manufacturing plant. Data from the supply chain to the warehouse, from the machine states to product quality and health and safety occurrences is recorded — somewhere. Usually in silos, rarely interacting with each other and sometimes recorded on paper! A lot of companies struggle with making use of the data that is collected, though there’s definitely an appetite at most firms to harness the power.

The first step to making data useful is to make it available. Make it available to different departments and staff so that they can make informed decisions. And that access should be as easy as possible, presented in an intuitive form.

The data in manufacturing is a gold mine, it has the insights an OEM needs to bring a step change in manufacturing. But it doesn’t just have to be the machine that provides the intelligence to extract the insight. In 2005, two amateur chess players armed with 3 PC’s defeated supercomputer and grandmasters. Individually their skills were inferior to their opponents but combined it had the human intuition and insight with computer predictions. This human machine symbioses, when built into the software, can allow staff to find the insights they are looking for. PayPal had struggled with fraud in the early days. There were too many for them to manually catch so decided to write software that could predict fraudulent transactions. The software was able to catch 80% of those transactions but not the 20% which were more sophisticated and evolved to avoid detection. So, they built analysis tools to allow humans to use their intuition to detect suspicious activity in the data.

To keep up with the pace of development in the automotive industry and keep ahead of the competition, manufacturers will need to get all the insight they can get from their data. Intuitive visualisation and analysis tools at every level of the company would empower the organisation to move forward together.

read original article here