The Impact of AI in Manufacturing: Unleashing Productivity

Companies are in a race to embrace digital technologies like artificial intelligence (AI). These technologies are critical enablers of the Fourth Industrial Revolution (also known as Industry 4.0) and will ultimately empower the manufacturing market to continue to be the backbone of the global economy. Artificial intelligence in manufacturing is bringing factories into the future.

Industry-wide, manufacturers are facing a range of challenges that make it difficult to speed production while still providing high-value and high-quality products to their customers. All the while, companies need to implement a digital infrastructure that positions them to fully embrace the skills and knowledge of their best assets — people.

The manufacturing industry today relies on automation just as much as people. But the factory of the future, which is a marriage of physical and digital capabilities, requires more: real-time data, connectivity and AI technology at the forefront. In fact, more than 80% of C-suite executives believe they must leverage AI to achieve their growth objectives.

The explosive growth of the electronics goods market means that there is little room for error or time to waste when embracing AI in manufacturing. Customer requirements for delivering on-time and on-budget product are of the utmost importance, and efficiency is a goal in everything manufacturing and supply chain management. AI’s ability to drive impact in this regard is real.

Manufacturing companies that adopt AI early will reap the biggest benefits. A McKinsey analysis projects a significant gap between companies that adopt and absorb artificial intelligence within the first five to seven years and those that follow or lag. The analysis suggests that AI adoption “front-runners” can anticipate a cumulative 122% cash-flow change, while “followers” will see a significantly lower impact of only 10% cash-flow change.

Think about injection molding machines. There are three parameters that affect the molding quality and product: the pressure on the injection, speed and temperature. At Jabil, we’ve been applying an AI solution and data analytics to analyze the parameters and track the temperature and pressure to detect common deviations.

High-resolution cameras with AI-based recognition software can perform quality checks at any point of the production process and help us accurately identify points where a product becomes defective. Is it because the machine isn’t functioning well? Or is it some other factor that is affecting the quality of the product? When we can answer these questions, the manufacturing processes become faster and more effective and produce higher quality products. This can be extremely beneficial for closely supervised industries like automotive and aerospace that must meet stringent quality standards set by regulatory agencies.

In fact, BMW Group already uses AI to evaluate component images from its production line, spotting deviations from quality standards in real-time. In the final inspection area at the BMW Group's Dingolfing plant, an AI application compares the vehicle order data with a live image of the model designation of the newly produced car. Model designations, identification plates and other approved combinations are stored in the image database. If the live image and order data don't correspond — for example, if a designation is missing — it sends a notification to the inspection team.

2. Augment Human Capabilities

The ultimate goal of artificial intelligence is to make processes more effective — not by replacing people, but by filling in the holes in people’s skills. By working side-by-side, the collaboration of people and industrial robots can make work less manual, tedious and repetitive, as well as more accurate and efficient.

To that end, Canon uses Assisted Defect Recognition — a combination of machine learning, computer vision and predictive analytics — to supplement human skills. The software examines manufacturing components with industrial radiography (X-ray) and images to determine the integrity of each part and its internal structure. With only a specialized technician, the examination process can be highly manual and error-prone. But with computer vision and machine learning, the Assisted Defect Recognition technology can analyze images of inspected parts, identify potential defects (including those that may be missed by the human eye), and learn and improve the technology’s accuracy based on human acceptance or corrections of the results.

One thing that we have been successful in doing at Jabil is deploying AI initiatives on natural language processing and learning. For instance, people need to pick up and identify the right trade compliance code to fill in when they do trade filing. In this task, accuracy is essential. If someone picks up the wrong commodity code and files it, that could result in picking up a dangerous good or a raw, hazardous good. We can now supplement the manual labor with artificial intelligence to pick up the right code so that we can file it properly.

3. Enable Preventative Maintenance

Almost 30% of use cases of AI in manufacturing are related to maintenance, per a Capgemini study. This makes sense considering that, in manufacturing, the greatest value from AI can be created by using it for predictive maintenance (about $0.5 trillion to $0.7 trillion across the world’s businesses).

Predictive maintenance analyzes the historical performance data of machines to forecast when one is likely to fail; limit the time it is out of service; and identify the root cause of the problem. Yield-energy-throughput (YET) analytics can be used to ensure that those individual machines are as efficient as possible when they are operating, helping to increase their yields and throughput and reduce the amount of energy they consume.

AI’s ability to process massive amounts of data, including audio and video, enables it to quickly identify anomalies to prevent breakdowns — whether that be an odd sound in an aircraft engine or a malfunction on an assembly line detected by a sensor.

With a machine failure, production stops. Meanwhile, predictive maintenance typically reduces machine downtime by 30-50% and increases machine life by 20-40%, according to a McKinsey article. With manufacturing’s increasing reliance on machinery and need to boost uptime and productivity, companies require much more than good luck and happy thoughts to keep production humming.

How to Successfully Implement AI in Manufacturing

The big challenge with AI implementation — which exists beyond manufacturing — is the abundance of data. You either don’t have enough data or you have so much that it becomes overwhelming and not actionable. In many manufacturing environments, most are still unable to extract certain data from machinery. Therefore, the AI is unable to highlight patterns and outliers. If you don’t know the process, it is very difficult to improve it.

Our governing principle in driving Industry 4.0 or smart factory initiatives is that, “If we are able to digitalize it, then we can visualize it.” After we can visualize it, we can optimize it.

There is abundance of data we generate in the manufacturing process and it is important we aggregate, catalog and use the data to solve the business problem. The definition of data and how we govern data is absolutely important. Data must be consistent, reusable, transparent, trustworthy and open. It is also important that we have a strategy on how we store and use data in the physical and logical perspective.

Data scientists are key to successfully incorporating AI into any manufacturing operation. They are needed to help companies process and organize the big data, turn it into actionable insight and write the AI algorithm to perform the necessary tasks.

But the data scientists themselves cannot do all the work. The business owners who understand the processes involved in manufacturing and production are familiar with how each parameter and factor affected will be influencing the outcome from the AI algorithm. Consequently, business involvement is vital.

Rolling out successful AI projects takes time. Think about AI as a brain; you need to train it. You probably need to have a process for the machine learning algorithm. We do need the process owner and the sponsorship of the management to know that this takes time. You will not see immediate effects; it’s a process.

Still, imagination is never-ending, and AI capabilities will be too. Think about our brains; they contain unlimited power. The AI evolution will be the same for manufacturing organizations. Productivity and efficiency will be rocketed to new heights, processes will be smoother and the future possibilities are endless.