10 use cases of AI in manufacturing (Part 1)

Accenture and Frontier Economics estimate that by 2035, AI-powered technologies could increase labor productivity by up to 40% across 16 industries, including manufacturing. In the same paper, the authors claim that AI could add an additional 3.8 trillion dollars GVA in 2035 to the manufacturing sector, which is an increase of almost 45% compared to business as usual.

Andrew Ng, the co-founder of Google Brain and Coursera, says: “AI will perform manufacturing, quality control, shorten design time, and reduce materials waste, improve production reuse, perform predictive maintenance, and more.”

And he’s correct. AI is already transforming manufacturing in many ways. Let’s have a look at some of the use cases of artificial intelligence for manufacturers.

Quality checks 

Some flaws in products are too small to be noticed with the naked eye, even if the inspector is very experienced. However, machines can be equipped with cameras many times more sensitive than our eyes – and thanks to that, detect even the smallest defects. Machine vision allows machines to “see” the products on the production line and spot any imperfections. The logical next step might be sending the pictures of said flaws to a human expert – but it’s not a must anymore, the process can be fully automated. Landing.ai, a company founded by Andrew Ng, offers an automated visual inspection tool to find even microscopic flaws in products. The system recognizes defects, marks them, and sends alerts.

Prediction of failure modes

Do you know the story about Abraham Wald and the missing bullet holes? And it’s a true story, may I remind you. Abraham Wald was a brilliant statistician. During World War II, he was asked by the Royal Air Force to help them decide where to add armor to their bombers. You don’t want your planes to be shot down, and neither adding too little armor nor adding too much of it works. The British analyzed the bombers that returned to Britain and found that most damage was done around the fuselage area of the bomber. Using simple reasoning, they should reinforce this part of the plane, right? They should not. The sample didn’t include the bombers that never made it home. And Wald was only looking for the “missing holes” – those around the engine. If a plane was shot there, it never came back. And the damage around the fuselage still didn’t stop the planes from returning to Britain. That’s were survival bias happens – we select some data to take into consideration and overlook other, often due to lack of its visibility. This can lead to false conclusions.

We can make false conclusions considering products and processes, too. Products can fail in a variety of ways, irrespective of the visual inspection. A product that looks perfect may still break down soon after its first use. Similarly, a product that looks flawed may still do its job perfectly well. The way we observe objects and flaws is biased and many things may be different than they seem. With vast amounts of data on how products are tested and how they perform, artificial intelligence can identify the areas that need to be given more attention in tests.

Predictive maintenance

Predictive maintenance allows companies to predict when machines need maintenance with high accuracy, instead of guessing or performing preventive maintenance. Predictive maintenance prevents unplanned downtime by using machine learning. Technologies such as sensors and advanced analytics embedded in manufacturing equipment enable predictive maintenance by responding to alerts and resolving machine issues. An excerpt from Deloitte’s The digital edge in life sciences report explains how IoT contributes to predictive maintenance:

An example of the use of Internet of Things and machine learning can be illustrated by predictive maintenance of machines used for manufacturing titanium implants. Titanium’s hardness requires tools with diamond tips to cut it. The level of dullness of the diamond tips, and thus the optimal time to sharpen them, has been difficult to figure out because of many different variables that affect it. The use of vibration or sound sensors and torque monitors can help assess the state of the machinery, as dull tips move and sound differently.

Predictive maintenance is already used by a number of manufacturers, including LG and Siemens. Roland Busch, Siemens AG CTO, says: “By analyzing the data, our artificial intelligence systems can draw conclusions regarding a machine’s condition and detect irregularities in order to make predictive maintenance possible.”

Since research conducted by Oneserve in the UK shows that 3% of all working days are lost annually due to faulty machinery, and the impact of machine downtime was estimated to cost UK manufacturers more than 180 billion pounds a year, predictive maintenance is gaining more popularity to help prevent losses.

Source: Neoteric