The key stages to deliver Artificial Intelligence into manufacturing production

Stanley Black and Decker are at the forefront of delivering digital solutions to industry and operate across manufacturing verticals, with a vision for digital transformation to bring in cutting-edge technology to solve the hardest problems.

The ambition is to build connected factories across the globe using technologies including: AI and machine learning; robotics; digital apps and digital twins.

Bala explained, “Among these technologies, AI is at the centre of building connected factories; but when we speak about AI, we are not only referring to deep learning AI, in fact it is much more than deep learning. It is machine learning, statistics and all those technologies that fall within the AI domain”.

The benefits of AI in manufacturing

According to Bala, the main reasons why machine learning has taken off falls into four categories: Data quantity; computing power; convolutional neural networks and marketing and media.

“The amount of data has grown exponentially. And the collection of data continues to proliferate in all aspects of business and society. In manufacturing, for example, we find vast amounts of data and we have got single instruments producing 40-dimensional data sets, so we’ve got a huge amount of data.

“There is also a huge amount of computing power with the element of the GPU, so we can either process in the cloud, or the edge, and computing power is very cheap. Combined with the development of neural networks, which has been a game changer for machine learning, there are new AI solutions opening up for manufacturers across the board.”

Machines “are already much better at doing certain tasks than humans” Bala noted and therefore it is simply a question of using the right technology tools to do the right jobs.

This is particularly evident in the field of pattern recognition and image classification – where machine learning algorithms have been outperforming humans for more than 15 years.

Where tasks are either too repetitive, or too hard, for humans – then these should obviously be assigned to AI, freeing up operatives to focus on where they retain a comparative functional advantage.

Bala noted that, “The most important use cases for industrial AI are to improve the accuracy, consistency and rate of throughput for manufacture.”

Delivering AI into manufacturing production

Software development is critical to the delivery of AI. You need to build the right infrastructure in order to carry out a successful implementation including, your own full specifications; development test production server stages; QA teams, and your own production teams that can deliver a 24/7 operation.

In addition you need a clear pathway and Bala suggested the stages set out in the flow chart below:

It is also important to build an ecosystem and ensure partnership building as Bala explained, “Thought leaders are in your area, they could be your suppliers, so you can essentially turn a supplier customer relationship into a strategic partnership.”

But how can a manufacturer ensure that the business truly engages with AI to maximise ROI? Bala suggested this is where automated machine learning comes in.

“One of the reasons I love automated machine learning is that it commoditises AI,” he concluded. “We can now build fully-functional AI models without requiring any AI expertise.”

This has the effect of significantly widening the scope for adoption, driven by the needs of the business. You will still need your software developers to assist in the process, but now a much greater pool of business experts can be involved. This will drive a lot more value of manufacturing AI deployments in future.

*Header image courtesy of Depositphotos