Photo credit: One Prod
When we have small rotating machinery, it will not be difficult to run corrective maintenance or preventive maintenance. But when we have large size of rotating machinery, performing corrective or preventive maintenance may not be so easy anymore.
Large machine will have many type of sections and components (bearings, gears, etc).
When the machine breaks down, it will take very long time to check all the sections and components one by one, just to find the part that causing the problem. And when the part is not available in our inventory, it will take even longer time to order the part, until it arrives.
Running a preventive maintenance for large machine may also reduce the machine uptime due to routine machine dis-assembly plus inspection of entire sections and components.
For such above case, where downtime period is no longer acceptable, we can shift from corrective maintenance or preventive maintenance strategy to predictive maintenance strategy based on machine health condition monitoring.
Based on the periodicity of the data collection, machine health condition monitoring can be classified into below categories:
Photo credit: One Prod
Periodical Monitoring
Periodical monitoring is done mostly by using portable data logger with at least one vibration and temperature sensor. Other than vibration and temperature sensor, we may need to use also the measure of the rpm of the machine. We may use either a tachometer or stroboscope. The data will be stored at the data logger and later can be uploaded to the maintenance server for further analysis. Since it is a portable data logger, it is commonly designed to have rugged structure, allowing to be used at harsh environment. Some manufacturers equip this device with special module for balancing. Some manufacturers also provide automatic diagnosis to help the maintenance engineer to sharpen and increase the accuracy of their diagnosis. The periodicity that commonly used for this type of condition monitoring is couple of weeks up to one month, depending on the failure characteristic and also cost effectiveness.
Semi-Online Monitoring
For periodical monitoring, since it uses portable data logger device, it may still need to send engineers to be present physically at the machinery area, to mount the sensors and collect the data. The challenges come when the machine is located at the hazardous area or area which is difficult to access. In this type of case, we can shift to semi-online monitoring which commonly uses wireless sensors. These sensors will not be mounted during data collection and linked to portable data logger like periodical maintenance, but it will be continuously mounted at the machine body, and wirelessly connected to a gateway which will automatically transfer the data to the maintenance server. The distance between the sensor to the gateway can go up to 50 m, or even more. Some manufacturers also provide repeater/extender to allow the sensors to be mounted even farther from the gateway. This repeater/extender will be placed in between gateway and sensor. This repeater/extender can be used also if there is any barrier in between gateway and sensor location. And since the sensor is commonly a wireless type, it will have battery inside. The manufacturer designs the structure of the sensor to be able to survive for long period at very harsh environment, such as ATEX Zone 0. For this type of condition monitoring, the periodicity of data collection commonly used is one data per day.
Real Time Monitoring
For semi-online monitoring, the sensor will be periodically active to collect data and to check if there is any parameter that goes beyond certain threshold. This type of monitoring is good for ordinary machine that continuously work at relatively constant load and RPM. But if we have special machine that works at variable load and RPM, the operation is triggered by other system (doesn't work continuously), and has many different additional process parameter with different type of sensors that need to be monitored, then we may want to consider using real time monitoring system. This type of condition monitoring will continuously monitor the machine condition, once there is machine health parameter that goes beyond the threshold, it will trigger an alarm, and collect data for the maintenance team for further analysis. The periodicity of data collection for this type of condition monitoring can go up to once per hour or even more. This type of condition monitoring uses data logger that can accept various type of sensors for monitoring various type of process parameter, such as RPM, pressure, current, airflow, etc.
Photo credit: One Prod
Below are some factors that need to be considered during a setup of condition monitoring system:
Photo credit: One Prod
Vibration sensor
How many vibration sensors to use and where to mount it will be depending on kinematics of the machine and the failure expected to be experienced by the machine. Most of the manufacturers will offer two type of vibration sensors: single axis and tri axis. We may consider using tri axis vibration sensor in area we expect to see significant vibration signal at all three directions, such as coupling joint. The method of mounting the vibration sensor may influence the vibration reading, especially at high frequency range. The best way to mount the vibration sensor will be stud mounting, since this mounting method will give us the highest resonance frequency of the vibration sensor. But in many cases, creating threaded hole at the machine may not be an option. In this type of case, using cementing stud will be an alternative option. Cementing stud may also work well at curved surface, especially when it is not easy to find flat surface for mounting the vibration sensor.
Monitoring parameters
Before setting up the acquisition strategy, what we may need to do first is creating the parameters (vibration and/or process parameters) to be linked to sensor reading. Most of manufacturers provide many types of overall parameters to be used, depending on the kinematics and the failure characteristics of the machine. Some overall parameters will be calculated by the data logger, and others will be calculated by the software at maintenance server. Other than overall parameters, we may be able also to record signal in time domain or frequency domain (spectrum). We may be able to run more detail analysis by deeply looking at these signals and checking the frequency content.
Data acquisition/collection strategy
After setting up the parameters, we may start to set our acquisition strategy. For real time monitoring, it will be very useful to set machine working "status/condition", based on the process parameters we created before, such as ON, OFF, Low Speed, High Speed, etc. These machine statuses will help us later to filter the data and make it easier to run more detail diagnosis. After setting up the machine status/condition, then we can set the acquisition periodicity. For real time monitoring, it will not be an issue to collect many data per day, but we may need to remember that short acquisition periodicity may lead to huge size of database. In this case, we may use feature provided by the maintenance software to automatically delete unnecessary data after certain period. For semi online monitoring, since sensor battery has lifetime. And this lifetime will be depending on how long the sensor is "active" every day. In order to save the lifetime of sensor battery, semi online condition monitoring may use low power wireless connection. Since sensor will be active only when it collects data and transferring the data to the maintenance server through the gateway, the acquisition periodicity will influence the lifetime of the sensor battery. In addition, when setting up the acquisition periodicity, we may need to consider also the duration needed to complete data transfer from sensor to maintenance server. The minimum periodicity should not be shorter than maximum time needed to complete the data acquisition process.
Photo credit: One Prod
