When we talk about Internet of Things (IoT), it is often in the context of the latest consumer device with a useless web service, like the IoT toothbrush or garden rake. These things make the news because they are relatable and a bit funny. What doesn’t make the news so much is real IoT, the sort that improves businesses and even protects lives. At HPE Discover, I spoke to people from two different industries whose businesses are being transformed by IoT. Curiously, both companies are using industrial IoT to optimize maintenance. Their production processes are already automated.
Disclosure: I attended HPE Discover in Las Vegas as a guest of HPE. My flight, accommodations, and most meals were paid for by HPE. HPE did not solicit this post, nor has it had any input on its content before publication.
First off, some context. What is industrial IoT, and what is industrial automation? Industrial systems have their own tools and standards for automation. The primary tool is a programmable logic controller (PLC). This is what turns valves and motors on and off in a manufacturing plant. The PLC controls a process and makes sure that product is produced. The overarching control for a factory or manufacturing site is usually a SCADA system, which is used to coordinate and control the various parts of the factory. Both the PLC and SCADA focus on making the product. They will react to a fault in the pumps, valves, pipes, or robots in the factory, but they are primarily for making the product. The value in industrial IoT seems to be in preventing faults: monitoring all those pieces of equipment and making sure they are operating correctly. Sensors are being retrofitted to existing manufacturing equipment to send data back to an analytics platform. The equipment usually has power but seldom has an Ethernet connection, so the sensors are wireless. The first stage of analytics is on-site, often physically alongside the manufacturing process to remove any external service dependencies.
One company I learned about was Texmark Chemicals; I had a great chat with Doug Smith, who is the CEO. Texmark is a relatively small petrochemical processing company in Texas. Its plant is made up of various tanks connected by pipes and pumps, plus some other magic that we didn’t discuss. The pumps are critical to the operation of the plant, so there is a team of “pump whisperers,” who make sure all the pumps are working and schedule maintenance to correct any problems. The pump whisperers make their rounds of every pump in the plant to ensure every pump is working correctly. Pumps are replaced on a schedule when their probability of failing is increasing. Many of the replaced pumps have years of life remaining; other pumps fail prematurely. A pump failure is a bad scenario. If a pump fails during production, then a whole batch will probably be ruined, at an enormous financial cost. Putting temperature sensors on every pump and getting continuous readings reduces the need to walk to each pump. An overheating bearing is a sure sign that a pump is on the way out. Even better, pumps can be replaced or repaired when they are showing evidence of impending failure. The net result is longer service from pumps and fewer failures. Both mean lowered costs and a more profitable and productive business.
Another company is involved in motor vehicle component manufacturing. It makes car doors in a very automated and mechanized production process; robots do a lot of the work. There are control systems that control the production of doors, so the IoT system is again aimed at plant maintenance: making sure the robots are healthy. Assembly robots are complex and mobile, so adding temperature sensors isn’t practical. It turns out that monitoring the current draw from the robot is a good indicator of robot health. Since a particular robot can be programmed to produce a variety of different products, you cannot simply monitor the current. Rather, the current profile over time is monitored and compared to the typical current profile to produce the various parts. If the profile matches the part that is being made, then the robot is in good condition. If the profile matches a different part or doesn’t match any part, then there is a problem. This is where the analytics part of IoT is important; the raw data isn’t enough. These analytics happen right on the manufacturing line. Some of the results are sent to central systems, but there are immediate analytics on-site.
Both of these industrial IoT use cases highlight another important factor: computing alongside the sensors. The IoT data needs to be analyzed close to the industrial process, and feedback needs to happen in near–real time. If a robot is malfunctioning or a pump is failing, the company cannot wait until the end of the week to plan maintenance. It also cannot afford for the analysis to wait due to an Internet outage. The processing needs to happen on-site with the industrial process.