Manufacturers are beginning to utilize artificial intelligence (AI) to support plant optimization. IoT sensors send data on the vibrations and temperature of machines to the cloud where AI algorithms compare the data to a growing database of machine signatures. Each machine has its own vibration and temperature signature. Based on this, the system can predict if a machine is going to break before it does. The system can make recommendations on what maintenance is required to prevent unplanned downtime.

Augury is a company that utilizes AI to analyze data from IoT sensors. “Our job is to make production lines more reliable. We offer a full turnkey system for machine health. That includes the sensors, connectivity, and the data algorithms,” Saar Yoskovitz, CEO of Augury, told Design News. “We use one type of sensor with different technologies, including vibration, magnetic, and temperature. It’s a holistic system for mechanical and electrical equipment such as drives and motors. It examines the health of the hardware using artificial intelligence software. The system covers 80% of the machines in the process industries.”

Part of the value of the system is its ability to communicate with plant personnel. “More than just diagnosing and sending the alert, the system makes sure you act on the alerts. The people on the production line receive plain English descriptions of the problem,” said Yoskovitz. “The alert explains what they need to do to fix the equipment. It also explains how much time the hardware has before it fails. For the more experienced technician, we provide a greater degree of analysis.”

Related:COVID's Impact on AI, IoT, Edge Computing, and Analytics

Reading the Needs of the Machines

The challenge for data analytics is its ability to manage the complexity of multiple machines. Each machine needs to perform optimally for the production line to run efficiently. “On the production lines, these machines are complex. You may have a motor that is dying, and you need to replace a bearing or add more oil. You may need to align the machine or open or close a valve. With the more complex machines, you need operational context,” said Yoskovitz. “This is where the AI is going. Maybe the equipment is vibrating because the viscosity is too high. Or maybe the raw material is different. The analytical tool helps the advanced engineer pinpoint where the issue is.”

For data analytics to function well is has to connect and share data from disparate systems. “The AI software analyzes the performance of each machine. The humidity outside may change the machine’s behavior. You have tools for production – MES and ERP – and you have tools for assets, the machine asset performance tools,” said Yoskovitz. “The link between these two tools is the predictive maintenance system. It can look at the same data as the control system. In the past, the systems have had competing incentives. We want to align the two on data.”

The Benefit is Productivity and Agility.

Beyond simply keeping the machines in decent running order, the data analytics also can analyze the optimal performance of the overall line. “Back in the 00s, sales and marketing looked at revenues. Now, revenue is seen as a lagging indicator. The leading indicators now may be the number of phone calls,” said Yoskovitz. “Likewise, we have tools to look at real-time indicators of production. Technology has changed how we work to be more productive. We’re using the data to bring greater efficiency.”

So much of machine health and overall value comes down to whether plant operators can predict performance. “When you look at the production lines, you’re looking at productivity and efficiency. Predictability is the key. Are the machines working today? Are they working all this week? Will they work next week?” said Yoskovitz. “If you have good data, you can be more agile. You can switch between lines easier. You can quickly recover from mistakes. Agility is key. The data gives you predictability and agility.”

Variability Becomes the Goal

A well-operating plant can do more than simply run according to plan. Predicability and agility can support variability. “When the factory can quickly turnover ques or recipes, you’ll be more productive. If you have 100 recipes for the product, predictability lets you correlate the mechanical aspects of production,” said Yoskovitz. “Maybe you can adjust the recipe to make it more efficient, or maybe you can use the data to determine what product line would work best for a particular recipe. These choices become an algorithm. We can connect to the operational data to pull in the results and offer high-level insight.”

The AI data analytics can become a template for optimal performance that can be transported to other plants, a kind of bast-practices in a box. “Data can be used to determine what additional assets you acquire. What assets will match your customer’s needs? We can use AI to replicate increased capacity and throughput,” said Yoskovitz. “If we can replicate it in one facility, we can do it at the other 20 facilities. Maybe you won’t have to build a new facility. Using data for internal benchmarking can be huge.”

Rob Spiegel has covered manufacturing for 19 years, 17 of them for Design News. Other topics he has covered include automation, supply chain technology, alternative energy, and cybersecurity. For 10 years, he was the owner and publisher of the food magazine Chile Pepper.