Five ways machine learning will transform manufacturing in 2021

Emerging developments that will become reality with machine learning part of everyday operations

By Ingo Mierswa February 2, 2021

Technology advances, such as complex robotic systems and artificial intelligence (AI), transformed manufacturing over the past decade and driven what’s commonly called the Industry 4.0 Revolution. COVID-19 has further accelerated this transformation for many manufacturers as more plant operations need to be run effectively, and in many cases, be monitored and managed remotely. Machine learning (ML) — a branch of artificial intelligence centered around creating computer programs that learn from experience and improve their decision-making ability over time — is increasingly important in many industries, and manufacturing is no exception.

Cheaper sensors and data storage, as well as the maturation of big data technology, has allowed manufacturers to capture vast amounts of data, and ML allows enterprises to derive actionable intelligence from said data, enabling smarter equipment, improved quality and increased productivity. As enterprises continue to use machine learning (ML) as part of their everyday operations, here are five ways manufacturing is expected to develop in the year ahead:

1. Widespread predictive maintenance

Typically, factories have relied on regular maintenance schedules dictated by usage or time to determine when a machine needs to be serviced — or have even waited until equipment breaks down to conduct maintenance. Leveraging AI, manufacturers are developing predictive maintenance models trained on historical data about what led up to past equipment problems to predict when machines need maintenance (see Figure). They can then fire an alert and the equipment can be repaired.

Because equipment is only shut down for repair when it’s actually needed, instead of according to a schedule, these models can save both time and money. Furthermore, as fast innovation cycles are shortening product lifecycles across most product categories, and dramatic shifts in customer expectations are shortening delivery lead times, manufacturers must be faster and more flexible with machine repairs and retooling.

While predictive maintenance isn’t new to 2021, expect to see a dramatic increase in its widespread adoption this year, with enterprises using data from connected devices such as sensors embedded into equipment to remove the guesswork from maintenance decisions.

2. Improving energy efficiency

Most factories today operate on a 24/7 schedule to maintain optimal efficiency, requiring large amounts of energy to keep things moving. By taking energy prices, equipment maintenance, labor costs and inventory into account, ML algorithms can schedule the perfect time to perform energy-intensive activities. As a result, enterprises can maximize cost savings by running the right processes at the right time.

This company has already seen this practice in play with a number of manufacturers, including a major global petrochemical manufacturer who came to RapidMiner looking to reduce its power consumption. The company was consuming $20 million a year in energy but wanted to cut costs and be more environmentally friendly. By deploying AI models that were easily adjustable in real time and worked on sensor data, RapidMiner was able to reduce its power consumption by 5%.

3. Guaranteeing product quality

Regardless of how optimized a manufacturing process is, every factory experiences product defects. Although there are various options for trying to correct them, flaws are still commonplace and treated as a cost of doing business. With ML, manufacturers can significantly reduce the possibility of error while optimizing quality control efforts.

Instead of relying on humans to visually inspect each product on an assembly line, image recognition and other types of ML models can be trained to analyze images and detect anomalies early in a product’s creation. As a result, factories can ensure they’re creating high-quality products while reducing waste.

4. Creating a safer workplace

Anyone who has worked in a factory has experienced thorough, annual health and safety trainings and knows the importance placed on proper use of safety gear. While these tools are critical for workplace safety, new technologies, like AI, can help further avoid risk, because accidents happen even when proper protocol is being followed.

Data analysis from ML can augment video surveillance systems to recognize potentially unsafe practices, including being used to identify overworked or tired employees before they operate heavy machinery. ML also can be used with sensor data to reveal important insights about safety-system performance. By relying on AI to sift through the thousands of data points generated every second by the Industrial Internet of Things (IIoT) and other connected devices, employers can get automatic alerts about potential dangers and thus create a safer workplace.

One Fortune 500 mining and chemical production firm was able to use machine learning to identify an unforeseen variable in its production process that commonly led to huge Environmental Health and Safety (EHS) risk factors. Using an ML model, built by process engineers, operators were able to keep the plant from going offline, avoid mountains of administrative paperwork and reduce its overall EHS risk by about 90%. The company estimates it avoided more than six incidents per year using ML.

5. Forecasting and responding to real-time consumer demands

Forecasting consumer demands can be a daunting task and is challenging to do perfectly. Thankfully, AI programs can be used to forecast demand with an unmatched level of sophistication and accuracy. Drawing from new and historical data, ML models can help businesses understand which factors drive demand and how enterprises can adapt to variables on the spot.

On the flip side, demand sensing lets businesses track fluctuations in demand, as well as consumer purchase behavior, in real time. By analyzing data from warehouses and point-of-sale systems, ML can identify significant changes in sales to ensure that supply is not outstripped by demand.

The advantages of ML systems for these processes, instead of relying on strict rules, have been highlighted by the COVID-19 pandemic. At the outset of the pandemic, as lockdowns began, consumption of, and thus demand for, products changed radically, leading to things like food and toilet paper shortages. Regularly scheduled deliveries based on what was typically needed weren’t able to keep up with changing behavior.

Wrapping up

Although we’re in the midst of the Industry 4.0 revolution, the manufacturing industry is only beginning to fully embrace digital transformation. Prior to 2020, quality, process optimization and operational expense reduction were key business drivers of this transformation. But now, amid a global pandemic, safety, remote worker enablement and information transparency are being added to that list. Advanced technologies based on AI and ML will continue to drive innovation and change how manufacturers think about problems and how to solve them, paving the way for a safer, more efficient and more profitable future.

Original content can be found at Plant Engineering.


Author Bio: Ingo Mierswa, PhD Is the founder and CTO of RapidMiner. He has been an industry-veteran data scientist since starting to develop RapidMiner at the Artificial Intelligence Division of the TU Dortmund University in Germany. He has authored numerous publications about predictive analytics and big data. As founder and CTO of RapidMiner, he is responsible for strategic innovation and technologies. RapidMiner has grown around 300% per year over the last seven years.