Time and usage based models do not accurately predict the end of life (EOL) of a component or machine as well as physical monitoring because there are too many variables the model cannot consider. For example, there is heat, humidity, external damage, improper machine usage, etc. that affects the life of a device. Arguably, a physics based model is better than a statistical one, and monitoring can be better than run-to-fail. Drawing that together using IoT provides a holistic view of the factory. All of that improves product quality, worker safety, and boosts productivity through reduced costs.
The goal of the manufacturer with monitoring is to provide better process stability and to provide data that can be used by decision support systems. It also helps keep the Overall Equipment Effectiveness (OEE) metric up.
IoT helps the manufacturer build a cloud based control loop system to bring the output of the plant closer to the control signal by comparing sensor measurements to reference values. And that system is a control loop, since it does more than just simple monitoring, by communicating with factory automation equipment.
Case in Point: Industrial Lubricants
Consider one item that needs to be monitored: industrial lubricants. Over time they degrade, increasing their viscosity and reducing their ability to protect ball, needle or roller bearings, slideways, rollers, gears and bushings. As the lubricant breaks down friction increases and the component begins to wear. If the drain interval is not set correctly the machine could break. Or it can cause production to go outside of tolerances. And then there is the issue of increased energy and fuel usage. And if the machines breakdown too often and there is no monitoring then engineers do not have all the data they need to determine why.
The Role of Sensors, Big Data and the IoT Cloud
Wireless sensors, big data databases, and the IoT cloud have made it easier to gather this data and assemble it. Flow, temperature, etc. can be measure with a thermocouple and transducer and wireless transmitter. This measures the fluid state by looking at electrochemical impedance. That transmitter uses a wireless, wired, or cellular connection to transmit data to the monitoring cloud. It does this using IoT devices and the IoT gateway, which is a type of network router.
Gathering data like this continuously is better than the point-in-time approach of statistical process control. The problems with statistical process control include false positives, the inability to detect trends, and using the just mean, variance, and standard deviation to flag events when those by are, by definition, based on percentages and not current conditions. The goal of the IoT approach should be to reduce these false positives and to flag quality issues earlier than would SPC.
For example, streaming IoT data on equipment vibration lets engineers study the harmonic signature of a bearing or shaft. This is better than sending a single signal to a control system because it can be plotted on a dashboard and fed into a machine learning algorithm.
Sensors can also monitor gas, air, or steam leaks. They can monitor calibration. They can assess supplier quality to build supplier ratings. Monitor component attributes (density, volume, roughness, weight, dimensions, etc.) as they move through the manufacturing process flags supplier quality and faulty materials. Linear position sensors use LIVRT technology measure movement. They can be to, for example, monitor die platen position in plastic molding machines or roller position.
All of this monitoring lets the manufacturer match the maintenance schedule with factors that factually indicate the need to do maintenance. It provides engineers with data they need to improve processes and diagnose failures. And it keeps painting machines, presses, dies, and pumps working within specs thus leading to improved product quality and process control.