Predictive Condition Based Maintenance – Does it Bring Any Significant Benefit?

Since the onset of industrial IoT systems, condition based maintenance or predictive maintenance gained momentum. Several big giants (IBM/Siemens/Microsoft/TCS/GE) proposed their IoT platform and data solution to push forward with the concept of predictive maintenance. Underlying promise has been to stop unplanned downtime and to reduce maintenance cost since for large percentage of time regular maintenance is done even when no maintenance is required.

Since 2015, lots of proof of concepts have been in place for condition-based monitoring (in MachineSense LLC, we have done several betas for condition-based monitoring all over the United States and learned significant insight since Feb. 2016). It will be worthwhile to review the “real” return of investment or benefit of condition-based monitoring. Additionally, in this blog, I would also share what we have learned from our condition monitoring beta in the plastic industry.

Please watch this short introductory video

Is there a real return on investment for Condition Based Monitoring or Predictive Maintenance?

Answer depends on A) how critical are the machine in production process and B) how easily the machine or failed component can be replaced C) how expensive is abrupt downtime.

Let me share my experience in this respect.

In the extrusion process, typically high horse powered VFD motors are used. They can range from 20-1000HP depending on torque and pull required in the process. As motor power exceeds 50 HP, no one stores the spare parts. In 300 HP+ range, replacement challenge goes far worse as the Motor vendors do not keep any spare part. Only if a spare order is placed, they machine it to specs.

Therefore, if there is an abrupt failure of a high horse power motor, down time can be as severe as between three weeks to three months. For critical machines like extrusion VFDs, we have seen very high interest among the plant owners to track the Motor health predictably.

Now consider low powered motors like 5HP or 10HP. A plant can get the spare overnight. A couple of hours of production downtime for many plants may not be significant unless they run at 100% capacity.

However, even for the small motors, there are three applicable scenarios where condition monitoring is valued.

When there are hundreds of motors in a series – obviously it’s an impossible proposition to maintain so many small motors regularly. Most of the factories run them to failure and keep spares. However, this is not a good solution for given the high number of motors/pumps, number of failures/disruptions can be too high and frequent.

Therefore, for a large number of small motors/pumps, predictive maintenance is highly advisable since otherwise, factory has to live with too many failures too frequently. And currently, they do.

If the material waste is significantly high – motors or pumps, however small it may be or how easy it may be to replace by spares in hand, can be of consequence if they are part of conveying system of the materials where materials if not processed due to line failures, can be wasted.

This is the case more with food and pharma than plastic. Since during material processing all the materials are heated & dried (or humidified) to an ambient condition, there can be a high level of wastage if the material conveying is shut down abruptly due to pump or motor failure.

For complex equipment, like Molding Machines, smaller motors can be part of Servo or Stepper motor systems. In this case, a spare may be available but to refit to the machine and retest it thoroughly takes longer time. So just having replacement parts, won’t solve the downtime rules.

Therefore, the value proposition of predictive maintenance or condition-based maintenance can be of varied magnitude depending on several factors as mentioned above.

Predictive Maintenance/ Condition Based Maintenance vs. Preventative Maintenance – which one is more effective and when?

Although predictive maintenance has been in the hype and attracted lots of fancy data science, preventative maintenance or identifying the cause that is leading to the death of the machine didn’t get much attention. In my experience, preventative maintenance provides the highest return on investment. I would like to share a short case study.

We had a beta site in a rural town of Pennsylvania, a state of the art plastic factory. We have noticed blower of a particular pump was getting destroyed in each six to twelve months where as those blowers must last for at least five years. It was causing loss to the OEM, producer of the pumps because they have also sold a warranty on those blower replacements.

On every occasion, we have seen the how quickly bearing health of the blower was trending to failure. In this case, predictive maintenance can tell when to replace the blower, but it is not enough to save the OEM from incurring a loss of warranty resulted from frequent blower replacement.

We also looked at the vacuum and vibration pattern of the pump. We have noticed the pump has been pulling more vacuum than it has been designed too. It was operating beyond its safety limit. Typically, to avoid exceeding the specifications, the plant must buy the pump that is appropriate for the application, but it is easier said than done in the industry which has lost most of its experienced manpower due to retirement and non-replacement.

Unfortunately, pumps or motors are seldom purchased accurately to fit the specification. In this case, however, the simpler solution was available. Experienced engineers from the OEM, advised the plant manager to put the pump in correct vacuum line so that Pump doesn’t need to pull the vacuum beyond its capability and die prematurely. This saved repeated failure of the blowers.

Typically, a machine dies prematurely due to three primary cause:

Feeding electrical line has bad power quality – this is number one killer of the motor. Poor power quality can be caused by several issues in the power distribution lines-such as the presence of too many vectors and dc drives, poor ground and uneven tapping of single phase lines from three phase lines. Interested users can use MachineSense Power Analyzer to diagnose power line quality for machines.

In all of the cases, high level of harmonics, phase imbalance and ground noise can be a killer to the motor because it heats up the stator coil beyond its safety limit. One will observe frequent motor failure if the power quality of the incoming line is poor.

Beating the machine beyond its specification – this is another cause of premature failure. MachineSense Component Analyzer (CA) detects the abnormal vibration or can sense the anomalous vibration resulting from operation beyond the safety limit. If a rotor is beaten beyond its capability, it will have an unstable plane of vibration and CA detects that instantly. In this case, the additional vertical pressure is created on ball-bearing and bearing-cages leading to faster erosion/defect of the bearing.

And when the machine is not maintained properly, such as when filters are not changed as specified, oil change doesn’t happen, or someone may pour poor quality of machine oil, etc. This is perhaps the most common cause of early failure, but the challenge here is to detect poor maintenance. In MachineSense Vacuum Pump Analyzer (VPA), poor maintenance is continually monitored and sent as an alarm.

This Video is a Great Resource on Unplanned Failure and Downtime on Vacuum Conveying Pumps

If a factory can systematically eliminate the causes of premature failures, they will see much less down time and more efficient operation. Just because one can track failure, doesn’t mean machines should be left out to fail as a frequent replacement has its own cost even if “unplanned downtime” cost can be eliminated by predictive maintenance.

Machines do not break so often-so why money spend on 24/7 predictive monitoring?

Failures may be a rare event, but when it happens, cost impact is catastrophic. But the question may arise why to buy a 24/7 system when maybe once in a year, it may come to rescue?

In MachineSense, we have used our predictive maintenance IoT system also to monitor process so that users can reap benefit on a regular basis. Because process issues happen on a daily basis. Using the same predictive maintenance system, users can extract a lot of useful information of the machines and the process:

Is the machine being continually abused? Meaning is the system being operated beyond its safe limit?

Is ambient temperature/humidity condition conducive to the machine and its control?

How much of time machine has been switched on? What percentage of the time machine has worked?

In the case of vacuum pump analyzer (VPA), users can also see detailed vacuum analytic which can help them to diagnose any issue in vacuum lines.

So a predictive maintenance system can be used for both preventative maintenance and process monitoring using advanced data science. This brings more benefit to the plant managers as some of the old machines may not have PLC and therefore he may not be knowing the productivity of those old machines.

This alleviates the concern that only predictive maintenance doesn’t give a full return on investment on sensor system purchased to track condition-based monitoring. Smart buyers must look for systems which track as many vital info they feel necessary or helpful for their plant operation.

All of the MachineSense products can help the plant operators in predictive maintenance (condition based monitoring), preventative maintenance (cause discovery) and process diagnosis.