With all the data, can all failures be predicted? Even with all the data, is it worth the trouble? When all this data is analyzed and processed, a set of potential points of failure can be predicted. In a loose sense, this is what Condition Monitoring can do, identify and predict potential faults in an industrial system.
For example, let’s say some SKF bearing, used under a particular load and speed, could fail after 160 hours of use. Somewhere within the 160 hours, a bearing starts to unnaturally wear. Sensors connected to the bearing housing could detect out of spec vibration from the damaged bearing. Or temperature of the bearing might be outside its acceptable range, lessening lifespan. Perhaps the bearing is spinning in excess of its rated RPM. Maybe the whole machine the bearing is in is shaking ─ periodically increasing the effective load. A Condition Monitoring system can take this data and make predictions on when this bearing will fail and provide a clear view of when a replacement should be made. This level of condition monitoring can help prevent unscheduled downtime, potentially saving thousands of dollars.
In a case study, conducted by Dynapar, showed how Dynapar’s On-Site Condition Monitoring System was retrofitted to a machine at a paper manufacturer’s plant and ultimately saved the manufacturer tens of thousands of dollars. Like in the hypothetical example above, a bearing was the main point of failure in contention. Dynapar’s system was installed on the machine in minutes, and it began to log data instantly. Turns out, the vibration data showed characteristics of an inner-race defect. A replacement was needed, but the machine it was installed inside ran almost 24/7. Each hour of downtime was worth $10,000 in loss.
The paper manufacturer decided that the bearing did need replacement, but the machine was functioning adequately enough to wait until routine maintenance was to be performed on it. When it was time for the regular maintenance, the faulty bearing was replaced with zero additional downtime for the machine.
With enough sensors, enough data, the right software with the right algorithms, predictions on a machine can span not just in the near future, but in the far future. Condition Monitoring can hypothetically project a maintenance schedule up until the point when the machine or device is useless. With enough historical data based on how the machine's been being used for the last six months or a year, a properly developed algorithm, or series of algorithms, can determine the maintenance schedule all the way up to the end-of-life of the machine. The predictions could potentially tell when a machine will no longer be suitable for a manufacturing process. A company can then start planning for future expenses. This sort of planning is called Predictive Maintenance. That paper manufacturer took their first steps towards using Condition Monitoring data to execute a Predictive Maintenance plan the same week they started using it.
Another advantage to a Condition Monitoring system is its use in tracking manufacturing defects. If over time, the parts start being produced out of spec, the Condition Monitoring data could show where issues might be. In injection molding machines, there could be a dozens of sensors on the mold. When parts start showing defect, a Condition Monitoring system will show changes to the mold over time. Perhaps it’s simply a temperature drop on the mold was causing defects. The data will show this and the temperature can be adjusted, preventing a run of bad parts. Having the machine sensor data from when the parts first started being produced correctly would give a baseline for the future.
Predicting failures and product defects with Condition Monitoring combined with a Predictive Maintenance plan is the best way to avoid future downtime and additional costs. However, there is a flip-side.
Condition Monitoring might seem like science fiction magic, and sensors everywhere is the future, but having that much technology has an obvious disadvantage. The number one problem companies may face with Condition Monitoring is cost.
The immediate reaction is often to place Condition Monitoring devices all over systems in operation, watching as many aspects of operation as one can afford. There could be thousands of sensors collecting data at any one time. This is where many companies see cost escalating. Not every point of a machine or system needs to be monitored. Assessing what is most critical to a system, taking into account repair history and common failure points, is the best way to sidestep excessive cost. Analyzing the cost impact of potential failures is key in deciding what to monitor.
Condition Monitoring and Predictive Maintenance is only that, predictive. It could be argued that replacing parts that have not failed is a waste of parts and labor. If Predictive Maintenance is a newly scheduled downtime, time could be added to that list of wastes. Keep the paper manufacturer’s $10,000 an hour loss during downtime in mind. If a maintenance team already has work to be performed, adding new Condition Monitored preventative tasks may delay other repairs they may need to do. This could lead to other issues. If outside contractors are brought in to do the maintenance, then cost is the disadvantage again.
Human error is still a big problem. Despite all the monitoring, when repairs happen, they typically are done by human beings. By adding repetitive Predictive Maintenance, error could come from the humans performing the repair or replacement tasks.
There is a less obvious cost to using Condition Monitoring devices, skilled personnel. These employees would be the ones monitoring the software, analysts, or even software engineers. Keep in mind, they are usually not the ones that would repair or perform maintenance on the machine or system.
Most importantly, not every failure can be predicted. Condition Monitoring does, however, give the best control over that unpredictability.
Have a story tip? Message me at: cabe(at)element14(dot)com