Maintaining machines residing in various manufacturing facilities can be an arduous task.

Factories are often in remote areas, and can be miles away from plant managers who may be unable to access the equipment on a day-to-day basis.

However, machine maintenance is crucial. The health of a machine deteriorating in pharmaceutical research and development facility, for example, could lead to the destruction of experiments resulting in millions of dollars in lost revenue.

Avoiding this challenge is the goal of startup Petasense.

“We’ve built an end-to-end wireless predictive maintenance system and machine learning software that will help you monitor the health of all kinds of machines that are found in plants and facilities,” explained Petasense co-founder Arun Santhebennur in an interview with R&D Magazine.

Santhebennur, along with the other founder Abhinav Khushraj, officially unveiled their company in June 2017 after operating under a temporary state of secretiveness for the first few years.

Essentially, the system works using Internet of Things (IoT) technology in a similar manner as a Fitbit. However, instead of tracking a person’s health via movement, it tracks the health of machines via vibrations.

“I would say 80 percent of failures of rotating machines are caused by things like imbalances, misalignment, looseness, and other factors. They all show all up in vibrations before they actually happen,” explained Santhebennur.

The product developed by Petasense can work for a variety of rotating equipment like motors, pumps, and gear boxes as well as HVAC equipment like compressors, chillers, and fans.

Setting a Baseline

Santhebennur explained the installation process is fairly easy, even for people who are not familiar with his technology.

Customers identify the machines they want to monitor. Petasense then recommends the best location for their sensors to be placed, in order to provide the best analysis for the vibrations.

Users then activate a mobile app, connect to Wi-Fi, and begin sending vibration data to the cloud.

Petasense’s own version of cloud software stores this information, and machine learning algorithms start providing information about the machine’s health.

This process occurs by setting a baseline for the machine, explained Santhebennur.

Within the first two weeks, the startup’s sensors sketch out a baseline for how a machine is behaving. Doing this gives operators insight into what constitutes typical performance for their device.

“In the past, without machine learning, what people would do is they would set standard level alarms for all fans at like 0.5 inches per second. It was not specific to select machines and was not based on the machine’s behavior,” said Santhebennur.

Any possible misstep from that generic baseline could trigger an alarm, requiring a plant manager to go check on the machine.

The difference with Petasense’s technology is that engineers can keep adjusting the baseline in the software and keep eliminating false positives in order to generate a higher level of customization.

Machinery found in these facilities can be old, but Santhebennur assured the integrity of the data obtained through this technology is accurate regardless of machine age. 

“We’ve manufactured thousands of sensors now that have gone through several iterations, factory calibration process, and a robust testing process, before each sensor goes out. It doesn’t matter if the machine are new or old as long as they are mounted properly so we can get a good sense of machine behavior’s baseline,” elaborated Santhebennur.

Industry applications

The startup’s product suite caters to clients in various industries like power generation, oil and gas, food and beverage, pharmaceuticals, and buildings and facilities.

Petasense has its service monitoring all the HVAC equipment including chillers, pumps, and compressors for some large pharmaceutical companies. The sensor and software combinations are also in use at oil and gas refineries tracking the health of some of these pumps in order to help employees avoid costly issues associated with their failure.

Furthermore, installing this end-to-end solution in combined cycle power generation plants is useful because these buildings contain hundreds of pumps and cooling tower fans. They possess both a gas and a steam turbine together to produce significantly more fuel than a traditional fan.

Operating these sensors in a plant like this is important because each component can perform a task like providing cooling water to parts of the plant in order to maintain a consistent temperature. Failing to ensure these tools are working can create hazardous conditions.

Understanding the Industrial IoT Space

The industrial internet of things (IoT) space is booming with large companies like IBM and GE, which are creating software for customers seeking a solution to enable smarter management of factories and similar surroundings.

Santhebennur explained how Petasense stands out, especially after announcing the company broke a recent milestone by logging 18 billion wireless vibration readings.

“It just shows the scale that's possible from a wireless perspective. Think about the old model. The old model was somebody would go and take readings once a quarter or once in six months. They would manually walk up to the machine and take a measurement. Now, with a wireless sensor, you could be doing that once every hour, so 24 times day instead of once in three months or six months,” he said.

Essentially, predictive maintenance is where Petasense feels it excels because this process generates a lot more data that can deliver a more precise analysis of the machine, thanks to the proprietary machine learning algorithm.

Overall, Santhebennur believes the future of the Industrial IIoT will see continued “explosion” of sensors to track more physical parameters and more breakthroughs on the wireless connectivity side.