predictive maintenance approach
12 Jun

Isn't predictive maintenance effective approach?

Something is wrong with Predictive Maintenance Solutions

If you are responsible for the maintenance and uninterrupted operations of factory assets then you probably know how brutal equipment downtimes can be. Data suggests that on average, one hour of downtime in a factory can cost $200K. This can be as high as $500K per hour depending on what the factory is producing. That is significant enough to give panic attacks to any factory owner.

In this post, we will cover the various maintenance techniques used by industries. We will talk about the advantages and challenges of each. We will also cover some of the reasons why Predictive Maintenance, used alone, may not be the best practice and what would be the best way to build a comprehensive maintenance strategy.

What are the common types of maintenance approaches?

1. Reactive Maintenance:

Reactive Maintenance, as the name suggests, is performed when a failure has already occurred. As a strategy, this is unavoidable no matter how well defined your maintenance strategy is because, despite all efforts, failures do occur. While this approach has glaring disadvantages such as surprise repair costs, shorter asset life expectancy and safety issues, it does have some advantages as well. A reactive maintenance approach is suitable when you are not operating any critical systems and have plenty of redundancies. This way the startup costs are much lower and due to a lower requirement for maintenance staff the company can offer competitive rates to their customers.


2. Preventive Maintenance:

Everyday life in factories and even in people‘s lives depends on many systems and machines. Some of them are mundane enough that we don’t notice until they fail, while others are very critical and require constant operation.

For mundane systems, the consequences of failure might just be an inconvenience. For such systems fixing the problems as they happen (Reactive Maintenance) is usually acceptable. As systems become more critical, regular preemptive maintenance is usually performed.

To that effect, industries have been relying on Preventive Maintenance techniques. Although the exact details of Preventive Maintenance vary based on the equipment, it involves periodic inspection, evaluation and repair/replacement of parts to prevent failure. Typical activities in a Preventive Maintenance schedule involve cleaning, oil changes, adjustments and sometimes an overhaul.

This determination to replace parts is either based on the condition of the parts after an inspection or just based on the manufacturer-recommended life cycle (in hours of operation). This sacrifices factory owner's ROI with the assurance that short-term loss is better than a catastrophic failure. As true as that might be, there is still a non-zero chance of failure. Which then begs the question, is sacrificing ROI for a non-zero chance of failure worth the cost if and when that eventuality happens? Clearly, this practice greatly mitigates the chances of failure but results in wasted resources.

Despite the risk of ROI, there are some obvious benefits to using Preventive Maintenance, such as:

·   Finite equipment operational life

·   Fewer equipment failures

·   Better visibility of maintenance schedules


Some of the software tools used to perform and manage Preventive Maintenance schedules include:

3. Computerized Maintenance Management System (CMMS)

Sometimes also referred to as an Enterprise Asset Management (EAM) system, is perhaps the most commonly used tool for Preventive Maintenance. At its heart, a CMMS is basically a software that maintains a database of your factory’s assets and information about the maintenance plans. A CMMS can be a one-stop-shop to locate and manage all your assets down to the last nut and bolt. It is also used to store user manuals for all the equipment that is inventoried in the system and can be used to schedule, manage, update and execute all kinds of scheduled maintenance plans, work orders, purchase orders and vendors with invoices. It can also be used to issue notifications to the right individuals in the system when a work order is created or scheduled maintenance is coming up.

However, like with any system that relies on human intervention, a CMMS is only as effective as the people responsible for using it.

4. Condition Monitoring Systems

These systems are different from CMMS because they rely on the actual condition of the asset rather than on predefined maintenance schedules. These systems are considered more cost-efficient because they focus maintenance activities to equipment that shows signs of deteriorating performance or upcoming failure thereby increasing the time between maintenance repairs. The indicators for performance can be gathered in multiple ways including visual inspections, periodic spot measurements and sometimes continue through the use of sensors.

The fundamental goal of Condition Monitoring Systems is to schedule maintenance when they spot an upcoming failure and not because a certain amount of time has elapsed since the last maintenance. This approach, however, has some of the same challenges that Predictive Maintenance Systems could suffer from, as we outline later in this post.

Predictive Maintenance:

In recent years Predictive Maintenance has generated a lot of interest both from maintenance solution providers and customers alike. Predictive Maintenance relies on sensor data to determine the probability of failure before it happens. Current solutions are leveraging Machine Learning models to analyze the various types of data generated by factory machines. This data is then used to determine the state of the machines and attempt to predict failures. This approach does yield promising results and new developments in the area of machine learning are making predictions even more accurate.

Predictive Maintenance solutions seem even more relevant considering the cost of failure for critical equipment. Imagine having to shut down production because one of the critical systems broke down unexpectedly and in between scheduled maintenance windows. The cost of failure is not just in terms of lost production but also involves various other costs such as:

·   Cost of pulling maintenance crew to work on this work order which in turn delays their current tasks.

·   If maintenance crew is unavailable or unable then it can be prohibitively expensive to involve outside contractors for repairs.

·   Parts for such critical equipment have to be maintained in the inventory in excess.

·   If parts are unavailable at the time then procuring them for this work order would also require an expensive shipment.

So, overall, the cost of failure is so prohibitive that factory owners would rather risk the loss of value in replacing parts with Preventive Maintenance than risk an unexpected failure.

When used optimally, Predictive maintenance can lead to significant cost savings with better predictability and of course increased operational times of the systems. We will discuss the optimal ways to use Predictive Maintenance later in this post.


How does Predictive Maintenance work?

All Predictive Maintenance starts with data collection. This is typically accomplished by attaching sensors to factory equipment and networking them to transmit this sensor data. This data is then analyzed for anomalies and the probability of failure is determined in real-time. The future health of equipment can then be predicted in order to enable on-demand maintenance.

There are three components to this approach:

1.  Sensor Hardware and Data Collection

There are various types of sensors available in the market that can sense machine data. The choice of sensor hardware depends on the machines and equipment that is being targeted for monitoring. At a high-level, these are some of the applications that these sensors are used for:

Induction Motors: Typical vibration measurements are done in the horizontal, vertical and axial direction on motor bearings. This data can be used to determine issues such as misalignment and imbalance in induction motors. One or more accelerometers are mounted on the motor housing either permanently or magnetically depending on motor accessibility. Data from these sensors can then either be collected manually by connecting a probe or uploaded wirelessly to the predictive maintenance system.

Cooling Towers and HVAC Systems: Cooling towers and HVAC systems are very critical in many industries. These cooling systems rely on large diameter fans that are responsible for moving outside air through the cooling equipment. Since these fans have very long blades spinning at high speeds, even a slight imbalance or misalignment can create high vibrations. Vibration monitoring is therefore very critical to determine if the vibrations are exceeding the acceptable thresholds. Warnings and even shutdowns need to be triggered when these vibrations levels are breached. Mechanical and/or electronic vibrations switches can be used to trigger shutdowns.

Shipping Containers: Freight companies often carry sensitive cargo in shipping containers that need to be handled with care during loading and unloading. Very often this cargo gets lost or damaged in transport. Logistics companies are relying on vibration monitoring to detect shock and tilts to prevent damage and maintain a record of these events. This makes it easier to settle insurance claims. Through the use of miniature accelerometers, GPS modules and data loggers they are now able to log shock and vibration data during the shipment cycle from the dock to dock.

So, what is the problem?

One of the common reasons suggested against using Predictive Maintenance is that replacing Preventive Maintenance with Predictive Maintenance makes the maintenance strategies reactive instead of proactive. It may sound counterintuitive, but the reasons do have some merit. However, as you will read in this post, the reason is not indicative of any problems with Predictive Maintenance as such but rather something else.

The argument against Predictive Maintenance is that monitoring for failures in isolation will lead to taking equipment off-line in isolation as well. What this means is that since these failures will tend to be random then the equipment will also be shut down at random. Of course, the decision for shutdown will be based on the severity of the failure and also on how critical the equipment is to the operations but as the equipment in the factory begins to age, the rate of such failures will also accelerate. Since these failure predictions will happen very often and randomly, this will put a strain on the inventory procurements as well because they will have to order and hold more spares. This will eventually lead to maintenance teams being in constant fire-fighting mode rushing from fixing one potential failure to next resulting in a very reactive maintenance strategy. On top of that if the maintenance team does not have the expertise to fix the equipment then hiring outside help on a rush order can be prohibitively expensive.

In summary, Predictive Maintenance does not fix problems but rather just reports the problems before they become a failure causing maintenance teams to rush from one repair to next.

Cost of Failure vs Prediction Error

Fundamentally, the expected output of any Predictive Maintenance solution is the accuracy of the Remaining Useful Life (RUL) of the equipment being monitored. This is also one of the most important considerations especially if you consider the cost of an incorrect prediction. Among all the assets in a plant, only a percentage of equipment is critical enough to warrant the use of Predictive Maintenance. These critical assets are expensive and often require expensive components for repair/maintenance work. These assets are also critical to the plant’s production and need to operate at full capacity with no breakdowns. Most other assets can be maintained with typical Preventive Maintenance solutions without much impact on cost or production.

Now imagine the cost of an incorrect prediction. Erring on the side of failure is obviously cost prohibitive at best and life-threatening at worst, however, under-predicting can lead to huge costs that will make it hard to justify Predictive Maintenance over Preventive Maintenance. In either case, the cost of erroneous predictions inadvertently leads to a Reactive Maintenance kind of cost.


How to use Predictive Maintenance effectively?

As we saw earlier, each approach has it’s own advantages and disadvantages. One thing we can agree on is that none of the three approaches is dispensable. Reactive Maintenance cannot be avoided because there will always be factors that are out of one’s control. While Predictive Maintenance has significant cost advantages, relying only on Predictive Maintenance will lead to a very reactive style of maintenance thereby negating the cost advantages.

An ideal and effective approach to maintenance would require all three maintenance types to be used in conjunction. Since not all assets and equipment in a plant are critical, the ideal approach would use Predictive Maintenance for the most critical assets, Preventive Maintenance for essential assets and Reactive Maintenance for non-essential assets that have redundancies.

Switch to a new and better way of maintaining your assets

Request demo