First we want to define what we mean by “Predictive Maintenance”. In the manufacturing industry Predictive Maintenance is used to detect faults in production machines before they can cause catastrophic failure and bring down or bottleneck your entire line. Predictive Maintenance accomplishes this by constantly monitoring sensors connected to your machinery and reporting this to a database or a cloud based system.
Predictive Maintenance has proven to have a high ROI based on increased up time, the ability to plan downtime, and the increased efficiency of maintenance operations. It is a key component in the Industry 4.0 push with manufacturers from all industries implementing solutions throughout the world.
For many years the basis of preventative/predictive maintenance was primarily based off of one of two approaches: Time Based Maintenance and Condition Based Maintenance. While neither of these two types of maintenance are going anywhere what Industrial IoT solutions provide is a more accurate and valuable version of them.
Time/Schedule based maintenance was the only real version of Predictive Maintenance and its foundation is based off of averages and assumption. Every X hours of operations we assume this machine may start to malfunction so we’ll schedule an inspection. This is only loosely predictive maintenance and is far outshined by more modern Predictive Maintenance procedures.
Previously Condition Based Maintenance relied on machine operators or repair technicians to notice an issue in the machine during routine operations. Hopefully they would catch the issue before it led to downtime, wasted materials, and reworks.
With the introduction of Industrial IoT sensors can now not only detect when a device is showing signs of fault/failure, but in some cases can tell repair technicians exactly what the problem is.
These benefits are especially apparent in equipment that integrates external systems like pneumatics. A single technician may not be familiar with all of the sensor data for the pneumatic system as well as the machine, but if the overarching system monitors both the machine and pneumatic systems it can correlate data between them and reach a more accurate conclusion faster.
This leads to:
There is a lot of buzz in the manufacturing about “Industry 4.0”. You’ll see presenters show endless graphs of data and talk about analytics, logistics, and “the future” of industry. It sounds big and impressive and really it is, but all of that starts with the same data you collect when you implement Condition Based Predictive Maintenance.
The end goal of Predictive Maintenance is to prevent malfunctions on a machine before they start to effect a machine’s outputs.
The basic concept is to collect important data about about your machines, store it, and inspect the data for unusual values. Whether you have something as simple as manually reviewing the data at the end of shift, a simple script that checks the sensor data for irregular values, or as complex as sending the data through a machine learning algorithm the basic concept is the same. Get sensor data, store it, and inspect it.
In the following sections we’ll talk about common sensor types to use to start implementing Predictive Maintenance in Manufacturing Environments.
Vibration monitoring is the single most important indicator for Predictive Maintenance in 99% of applications. This is the data that will tell you when something is wrong, how bad the issue is, and also indicates what the issue most likely is. The best part is you don’t need machine learning or complex algorithms.
In our latest generations of vibration sensor we do a lot of the work for you in analyzing this data to give you the most pertinent information used in Predictive Maintenance applications.
To put the concept simply each motor has a speed that it rotates. Let’s say this speed is 100Hz which means the engine will cycle 100 times per second. This means if we look at our processed vibration data we should see a peak in vibration at ~100Hz. This tells us that the bulk of the vibration we’re seeing is happening at the expected frequency. Now imagine that data is showing two peaks: one at 100Hz and another at 400Hz. We know the motor isn’t operating at two speeds simultaneously so this second peak means there is an issue with the balance, mounting, alignment, or bearing of the motor.
You can also look at the amount of vibration the machine has. The velocity and acceleration are key indicators of machine health all on their own and are easy to monitor in any Predictive Maintenance procedure.
The accurate detection of exact issues does require technical knowledge. Most of the work in determining exact issues is standardized and covered in ISO 20816:2016.
Some motor are more complex than others due to things like variable speed motors which can change peak frequencies in the data. For those you will want to look at additional metrics as outlined below to flesh out your Predictive Maintenace approach.
The amount of current that a motor/machine is drawing is a key indicator in Predictive Maintenance for Manufacturing equipment. It is also extremely useful to monitor more complex machines with multiple motors or variable speed motors.
The general concept is fairly straightforward. A faulty motor will draw more power to try and do its job than a motor that isn’t having issues. Simply look for current draw outside the standard current draw of the motor and you can detect machine faults before total/catastrophic failure.
Current monitoring is best used in conjunction with other metrics like Vibration and Temperature to give your team a complete overview of machine health.
When it comes to monitoring machine health for Predictive Maintenance temperature is a very important aspect. It can monitor ancillary and environmental factors that can impact machine health.
For instance let’s say you have a machine that requires forced air cooling and the cooling system goes down. By the time you can see variances in vibration or current your machine has already overheated. High temperatures can compromise gaskets, bearings, lubricants, and many other components of a machine. By monitoring the temperature of the machine you can detect these failures before they impact your other metrics. This is the key aspect of a fully operational Predictive Maintenance application.
Temperature, Current, and Vibration are the most ubiquitous ways to implement a Predictive Maintenance application and monitor for abnormal behavior in almost any industrial equipment. Your Predictive Maintenance journey doesn’t have to start with integrating machine learning or complex cloud infrastructure in order to start being useful. As long as you are storing the data you can move to more complex Predictive Maintenance Procedures as your needs dictate.
If the motor of your lathe or extruder is getting too hot it can indicate a problem. If left unchecked this problem can cause damage to the unit and extended downtimes repairing the machine and waiting for replacement parts.
Now if that motor’s temperature was highly variable due to the environmental factors of your facility, variability in the machine’s raw material inputs, or even within the machine’s enclosure you may need another data source. This is where current comes in. If the machine starts drawing more power than usual AND the temperature of the motor is high you can more accurately predict that there is an issue with the motor. It is working harder than expected and expending excess energy in the form of heat.
The third data point, vibration, can not only be used to detect anomalous behavior of the machine before temperature and current variations can be detected, but can even be used to predict what part of the machine needs maintenance. Comparing this data against the guidelines outlined in ISO 20816:2016 you can easily use the vibration data to pinpoint issues such as bad bearings, faulty motor housing, loose bolts, or an out of spec drive rod.
NCD offers many sensor solutions to monitor exactly this kind of data. We even offer a single sensor that has all three types built in using our Industrial IoT Wireless Predictive Maintenance Sensor. This sensor is designed to be a one size fits most product. However there are going to be application that fall outside the bounds of the sensor’s capabilities, usually in the form of Current Monitoring. If you are wanting to get started collecting data for your move to IIoT you can always feel free to contact us. We have a team of highly skilled Hardware and Software engineers that can help you look at your application and find the best product and path based on your needs and resources.